‫ﺟﺎﻣﻌﺔ اﻟﻤﻠﻚ ﺳﻌﻮد‬

‫آﻠﻴﺔ اﻟﻌﻠﻮم‬
‫ﻗﺴﻢ اﻹﺣﺼﺎء وﺑﺤﻮث اﻟﻌﻤﻠﻴﺎت‬

‫ﻃﺮق اﻟﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ‬
‫) اﻟﺠﺰء اﻷول(‬

‫ﺗﺄﻟﻴﻒ د‪ .‬ﻋﺪﻧﺎن ﻣﺎﺟﺪ ﻋﺒﺪاﻟﺮﺣﻤﻦ ﺑﺮي‬
‫أﺳﺘﺎذ اﻹﺣﺼﺎء وﺑﺤﻮث اﻟﻌﻤﻠﻴﺎت اﻟﻤﺸﺎرك‬

٢

‫ﺑﺴﻢ اﷲ اﻟﺮﺣﻤﻦ اﻟﺮﺣﻴﻢ‬
‫اﻟﺤﻤﺪ ﷲ رب اﻟﻌﺎﻟﻤﻴﻦ واﻟﺼﻼة واﻟﺴﻼم ﻋﻠﻰ اﺷﺮف ﺧﻠﻖ اﷲ ﺳﻴﺪﻧﺎ وﻧﺒﻴﻨﺎ ﻣﺤﻤﺪ‬
‫وﻋﻠﻰ ﺁﻟﻪ وﺻﺤﺒﻪ وﺳﻠﻢ‪.‬‬
‫أﻣﺎ ﺑﻌﺪ‪.‬‬
‫هﺬﻩ هﻲ اﻟﻤﺴﻮدة اﻷوﻟﻰ ﻟﻜﺘﺎب ﻃﺮق اﻟﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ ﻟﻄﻼب ﻣﺮﺣﻠﺔ اﻟﺒﻜﺎﻟﻮرﻳﻮس‪.‬‬
‫هﺬا اﻟﻜﺘﺎب ﺳﻴﻈﻞ ﻣﺴﻮدة إﻟﻰ ﻣﺎﺷﺎء اﷲ ﻷﻧﻲ وﺑﺈذن اﷲ ﺗﻌﺎﻟﻰ ﺳﻮف أﻗﻮم ﺑﺘﻄﻮﻳﺮﻩ‬
‫وﺗﺠﺪﻳﺪﻩ وﺗﺤﺴﻴﻨﻪ ﺑﺸﻜﻞ ﻣﺴﺘﻤﺮ وﺳﻴﻈﻞ ﺑﺸﻜﻠﻪ اﻹﻟﻜﺘﺮوﻧﻲ هﺬا ﻷﻧﻲ أﻋﺘﻘﺪ ان اﻟﻌﻠﻮم‬
‫واﻟﺘﻘﻨﻴﺔ ﺗﺘﻄﻮر ﻳﻮﻣﻴﺎ وﺑﺸﻜﻞ ﻣﺘﺴﺎرع ﺑﺤﻴﺚ ان وﺿﻌﻬﺎ ﻓﻲ آﺘﺎب ﺟﺎﻣﺪ ﺳﺘﺎﺗﻴﻜﻲ‬
‫ﻻﻳﺘﻨﺎﺳﺐ ﻣﻊ دﻳﻨﺎﻣﻴﻜﻴﺔ اﻟﻤﻮﺿﻮع وﺧﺎﺻﺔ ﻓﻲ ﻋﺼﺮ ﺛﻮرة اﻟﻤﻌﻠﻮﻣﺎت واﻹﻧﺘﺮﻧﺖ‪.‬‬
‫ﻳﻐﻄﻲ اﻟﺠﺰء اﻷول ﻣﻦ اﻟﻜﺘﺎب اﻷﺳﺎﺳﻴﺎت اﻷوﻟﻴﺔ ﻟﻠﻤﻮﺿﻮع وﻳﺘﻄﺮق إﻟﻰ ﻣﻮﺿﻮع‬
‫اﻟﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ ﺑﺈﺳﺘﺨﺪام ﻧﻤﺎذج ‪ ARIMA‬واﻟﺘﻲ آﺎﻧﺖ اول ﻣﻌﺎﻟﺠﺔ رﻳﺎﺿﻴﺔ ﺟﺎدة‬
‫وﻣﺤﻜﻤﺔ ﻟﻠﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ ﺑﺈﺳﺘﺨﺪام اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ‪ Time Series‬آﻤﺎ ﺗﻄﺮﻗﺖ‬
‫ﻓﻲ ﺁﺧﺮ اﻟﻜﺘﺎب إﻟﻰ ﺑﻌﺾ اﻟﻄﺮق اﻟﺘﻘﻠﻴﺪﻳﺔ اﻟﻬﻮرﺳﺘﻴﻜﻴﺔ ﻟﻠﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ وﻓﻲ ﺟﻤﻴﻊ‬
‫أﺟﺰاء اﻟﻜﺘﺎب ﻗﻤﺖ ﺑﺘﻮﺿﻴﺢ اﻷﻣﺜﻠﺔ واﻟﺤﺎﻻت اﻟﺪراﺳﻴﺔ ﺑﺈﺳﺘﺨﺪام اﻟﺤﺰﻣﺔ اﻹﺣﺼﺎﺋﻴﺔ‬
‫‪ Minitab‬وهﻲ ﺑﺮاﻣﺞ ﺣﺎﺳﺐ ﻃﻮرت ﺧﺎﺻﺔ ﻟﺘﻌﻠﻴﻢ ﻋﻠﻢ اﻹﺣﺼﺎء ﺑﺠﻤﻴﻊ ﻓﺮوﻋﻪ‬
‫وهﺬﻩ اﻟﺤﺰﻣﺔ ﻣﺘﻮﻓﺮة ﻟﻠﻄﻼب ﺑﺎﻟﻤﺠﺎن‪.‬‬
‫اﻟﺠﺰء اﻟﺜﺎﻧﻲ ﻣﻦ اﻟﻜﺘﺎب وﻣﻮﺟﻪ ﻟﻄﻼب اﻟﺪراﺳﺎت اﻟﻌﻠﻴﺎ ﺳﻮف ﻳﺘﻄﺮق ﺑﺈذن اﷲ‬
‫ﻟﻤﻮاﺿﻴﻊ ﻣﺜﻞ ﺗﺤﻠﻴﻞ اﻟﺘﺪﺧﻞ ‪ Intervention Analysis‬وﻧﻤﺎذج داﻟﺔ اﻟﺘﺤﻮﻳﻞ‬
‫‪ Transfer Function Models‬وﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﺘﻌﺪدة‬
‫‪ Multivariate Time Series Models‬وﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺟﻬﻪ‬
‫‪ Vector Time Series Models‬وﻧﻤﺎذج ﻓﻀﺎء اﻟﺤﺎﻟﺔ وﻣﺮﺷﺢ آﺎﻟﻤﻦ ‪State‬‬
‫‪ Space Models and Kalman Filtering‬وﻧﻤﺎذج اﻟﺤﺪ ‪Threshold Time‬‬
‫‪ Series Models‬وﻧﻤﺎذج ‪ ARCH‬وﻧﻤﺎذج ‪ GARCH‬وﺗﻄﺒﻴﻘﺎﺗﻬﺎ ﻓﻲ اﻟﺘﻨﺒﺆ‬
‫اﻟﻤﺎﻟﻲ ‪ Finantial Time Series Forecasting‬آﻤﺎ ﺳﻨﺘﻄﺮق إﻟﻰ اﻟﺸﺒﻜﺎت‬
‫اﻟﻌﺼﺒﻴﺔ ‪ Neural Networks‬وإﺳﺘﺨﺪاﻣﻬﺎ ﻓﻲ اﻟﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ‪.‬‬
‫هﺬا وارﺟﻮا ﻣﻦ اﷲ ان ﻳﻮﻓﻘﻨﻲ ﻓﻲ إﻧﺠﺎز هﺬا اﻟﻌﻤﻞ ﻟﻮﺟﻬﻪ اﻟﻜﺮﻳﻢ وﻹﺛﺮاء اﻟﻤﻜﺘﺒﺔ‬
‫اﻟﻌﺮﺑﻴﺔ اﻟﻔﻘﻴﺮة إﻟﻰ ﻣﺜﻞ هﺬا اﻟﻜﺘﺎب‪.‬‬
‫ﺳﻴﻜﻮن هﺬا اﻟﻜﺘﺎب ﻣﺠﺎﻧﻲ ﻷي ﻃﺎﻟﺐ ﻋﻠﻢ وهﻮ ﺳﻴﻜﻮن ﻣﺘﻮاﺟﺪ ﻋﻠﻰ ﺷﺒﻜﺔ اﻹﻧﺘﺮﻧﺖ‬
‫ﻓﻲ اﻟﻤﻮﻗﻊ ‪http://www.abarry.net/or/or٢٢١book١.pdf‬‬
‫واﷲ اﻟﻤﻮﻓﻖ‪.‬‬
‫اﻟﻤﺆﻟﻒ‬
‫د‪ .‬ﻋﺪﻧﺎن ﻣﺎﺟﺪ ﻋﺒﺪ اﻟﺮﺣﻤﻦ ﺑﺮي‬
‫ﺟﺎﻣﻌﺔ اﻟﻤﻠﻚ ﺳﻌﻮد‬
‫ذو اﻟﻘﻌﺪة ‪ ١٤٢٢‬هـ‬
‫ﻳﻨﺎﻳﺮ ‪ ٢٠٠٢‬م‬
‫‪٣‬‬

‫اﻟﻤﺤﺘﻮﻳﺎت‬
‫ﻣﻘﺪﻣﺔ‬
‫‪ -١‬اﻟﻔﺼﻞ اﻷول‪ :‬ﻣﻘﺪﻣﺔ وﺗﻌﺎرﻳﻒ‪١٠..................................................................‬‬
‫‪ ١-١‬أﻣﺜﻠﺔ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ‪١٠.......................................................‬‬
‫‪ ٢-١‬اﻟﻐﺮض ﻣﻦ دراﺳﺔ وﺗﺤﻠﻴﻞ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ‪١٠ .................................‬‬
‫‪ ٣-١‬اﻟﺨﻄﻮات اﻟﻤﺘﺨﺬة ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ ‪١٠ . ................................................‬‬
‫‪ ١-٣-١‬ﺗﻌﻴﻴﻦ اﻟﻨﻤﻮذج ‪١٠ .. ...................................................................‬‬
‫‪ ٢-٣-١‬ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج ‪١١ ...................................................................‬‬
‫‪ ٣-٣-١‬ﺗﺸﺨﻴﺺ وإﺧﺘﺒﺎر اﻟﻨﻤﻮذج ‪١١......................................................‬‬
‫‪ ٤-٣-١‬ﺗﻮﻟﻴﺪ اﻟﺘﻨﺒﺆات ‪١١.....................................................................‬‬
‫‪ ٥-٣-١‬إﺳﺘﺨﺪام اﻟﺘﻨﺒﺆات ووﺿﻊ اﻟﻘﺮارات ‪١١ ...........................................‬‬
‫‪ ٤-١‬ﺗﻌﺎرﻳﻒ وﻣﺒﺎدئ أوﻟﻴﺔ ‪١١.....................................................................‬‬
‫‪ ١-٤-١‬ﺗﻌﺮﻳﻒ ﻣﺎﺿﻲ أو ﺗﺎرﻳﺦ اﻟﻈﺎهﺮة ‪١١ .............................................‬‬
‫‪ ٢-٤-١‬ﺗﻌﺮﻳﻒ اﻟﺤﺎﺿﺮ أو اﻵن ‪١١ .........................................................‬‬
‫‪ ٣-٤-١‬ﺗﻌﺮﻳﻒ أﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ ‪١١ .........................................................‬‬
‫‪ ٤-٤-١‬ﺗﻌﺮﻳﻒ أﺧﻄﺎء اﻟﺘﻨﺒﺆ ‪١٢ .............................................................‬‬
‫‪ ٥-٤-١‬ﺗﻌﺮﻳﻒ اﻹﺳﺘﻘﺮار ‪١٢ .................................................................‬‬
‫‪ ٦-٤-١‬ﺗﻌﺮﻳﻒ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪١٢ .........................................................‬‬
‫‪ ٧-٤-١‬ﻣﺜﺎل ‪ : ١‬اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ‪١٢ ...................................................‬‬
‫‪ ٨-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﻐﺎﻳﺮ اﻟﺬاﺗﻲ ‪١٣ ....................................................‬‬
‫‪ ٩-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ‪١٣ ...................................................‬‬
‫‪ ١٠-٤-١‬ﻣﺜﺎل ‪ : ٢‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪١٣ .........................‬‬
‫‪ ١١-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ‪١٤ ......................................‬‬
‫‪ ١٢-٤-١‬ﻣﺜﺎل ‪ :٣‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪١٥..................‬‬
‫‪ ١٣-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻌﻴﻨﺔ ‪١٦.........................................‬‬
‫‪ ١٤-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ ‪١٧ ..............................‬‬
‫‪ ١٥-٤-١‬ﻣﺜﺎل ‪ :٤‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ ‪١٨ .........‬‬
‫‪ -٢‬اﻟﻔﺼﻞ اﻟﺜﺎﻧﻲ‪ :‬ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪ ARMA‬وإﺳﺘﺨﺪاﻣﺎﺗﻬﺎ ﻓﻲ‬
‫اﻟﺘﻨﺒﺆ ‪٢٢................................................................................................‬‬
‫‪ ١-٢‬ﺗﻌﺮﻳﻒ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪٢٢ ........ (p,q‬‬
‫‪ ٢-٢‬ﺗﻌﺮﻳﻒ ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻟﺨﻠﻔﻲ ‪٢٢.........................................................‬‬
‫‪ ٣-٢‬ﺗﻌﺮﻳﻒ ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻷﻣﺎﻣﻲ ‪٢٢ .....................................................‬‬
‫‪ ٤-٢‬ﺗﻌﺮﻳﻒ ﻋﺎﻣﻞ اﻟﺘﻔﺮﻳﻖ ‪٢٢..................................................................‬‬
‫‪ ٥-٢‬ﺗﻌﺮﻳﻒ ﻋﺎﻣﻞ اﻟﺘﺠﻤﻴﻊ ‪٢٢ .................................................................‬‬
‫‪ ٦-٢‬أﻣﺜﻠﺔ ‪٢٣ .....................................................................................‬‬
‫‪ ٧-٢‬ﺧﺼﺎﺋﺺ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪٢٤ .............................‬‬
‫‪ ١-٧-٢‬ﻧﻤﻮذج )‪٢٤ ........................................................... ARMA(٠،٠‬‬
‫‪ ٢-٧-٢‬ﻧﻤﻮذج )‪٢٧ .................................................................................. AR(١‬‬
‫‪ ٣-٧-٢‬ﻧﻤﻮذج )‪٣١ .................................................................... AR(٢‬‬
‫‪٤‬‬

‫‪ ٤-٧-٢‬ﻧﻤﻮذج )‪٣٦ ..................................................................... MA(١‬‬
‫‪ ٥-٧-٢‬ﻧﻤﻮذج )‪٣٩ .................................................................... MA(٢‬‬
‫‪ ٦-٧-٢‬ﻧﻤﻮذج )‪٤٠ ............................................................. ARMA(١،١‬‬
‫‪ ٧-٧-٢‬ﺧﻮاص ﻧﻤﺎذج )‪٤٧ ................................................ ARMA(p,q‬‬
‫‪ -٣‬اﻟﻔﺼﻞ اﻟﺜﺎﻟﺚ‪ :‬ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ‪٤٩ ......................................‬‬
‫‪ ١-٣‬ﻋﺪم اﻹﺳﺘﻘﺮار ﻓﻲ اﻟﻤﺘﻮﺳﻂ ‪٤٩ ..............................................................‬‬
‫‪ ٢-٣‬ﻋﺪم اﻹﺳﺘﻘﺮار ﻓﻲ اﻟﺘﺒﺎﻳﻦ ‪٥٠ ................................................................‬‬
‫‪ ٣-٣‬ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪٥٢ ...... (p,d,q‬‬
‫‪ ١-٣-٣‬ﻧﻤﻮذج )‪٥٢ ....................................................... ARIMA(١،١،٠‬‬
‫‪ ٢-٣-٣‬ﻧﻤﻮذج )‪٥٢ ........................................................ ARIMA(٠،١،١‬‬
‫‪ ٣-٣-٣‬ﻧﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ﺑﺈﻧﺠﺮاف ‪٥٣ .................................................‬‬
‫‪ ٤-٣‬داﻟﺔ اﻷوزان ) ‪ y (B‬وﺗﻤﺜﻴﻞ ﻧﻤﺎذج )‪٥٣ ............................... ARMA(p,q‬‬
‫‪ ٥-٣‬اﻣﺜﻠﺔ ﻟﺪاﻟﺔ اﻷوزان ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ‪٥٤ .....................................................‬‬
‫‪ ١-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٤ .................................................... AR(١‬‬
‫‪ ٢-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٥ .................................................... MA(١‬‬
‫‪ ٣-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٥ .................................................... AR(٢‬‬
‫‪ ٤-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٦ ................................................... MA(٢‬‬
‫‪ ٥-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٦ ........................................... ARMA(١،١‬‬
‫‪ ٦-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٧ ................................................... ARI(١‬‬
‫‪ ٧-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ )‪٥٧ .................. ARIMA(١،٠،١‬‬
‫‪ ٦-٣‬ﺑﻌﺾ ﺧﻮاص داﻟﺔ اﻷوزان ) ‪٥٨ .................................................... y (B‬‬
‫‪ -٤‬اﻟﻔﺼﻞ اﻟﺮاﺑﻊ‪ :‬اﻟﺘﻨﺒﺆات ذات ﻣﺘﻮﺳﻂ ﻣﺮﺑﻊ اﻟﺨﻄﺄ اﻷدﻧﻰ ﻟﻨﻤﺎذج )‪٥٩ ....... ARMA(p,q‬‬
‫‪ ١-٤‬ﻧﻈﺮﻳﺔ ‪ :٢‬أﺧﻄﺎء اﻟﺘﻨﺒﺆ ‪٦١ .........................................................................‬‬
‫‪ ٢-٤‬ﻣﺠﻤﻮﻋﺔ اﻟﻤﻌﻠﻮﻣﺎت ‪٦٢ ................................................. Information Sets‬‬
‫‪ ٣-٤‬ﻧﻈﺮﻳﺔ ‪ :٣‬اﻟﻤﺘﻨﺒﺊ ذا ﻣﺘﻮﺳﻂ ﻣﺮﺑﻊ اﻟﺨﻄﺄ اﻷدﻧﻰ ‪٦٢ ............................................‬‬
‫‪ ٤-٤‬ﻗﺎﻋﺪة ‪٦٢ ............................................................................................. ٢‬‬
‫‪ ٥-٤‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﻨﺒﺆ ‪٦٢ .................................................................................‬‬
‫‪ ٦-٤‬دوال اﻟﺘﻨﺒﺆ ﻟﻨﻤﺎذج )‪٦٢ ....................................................... ARIMA(p,d,q‬‬
‫‪ ١-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٣ ............................................................... AR(١‬‬
‫‪ ٢-٦-٤‬ﺷﺮط اﻹﺳﺘﻤﺮار ‪٦٣ ............................................................................‬‬
‫‪ ٣-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٤ .............................................................. AR(٢‬‬
‫‪ ٤-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٥ ................................................. ARIMA(٠،١،١‬‬
‫‪ ٥-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٦ .............................................................. MA(١‬‬
‫‪ ٦-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٦ .............................................................. MA(٢‬‬
‫‪ ٧-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٧ ..................................................... ARMA(١،١‬‬
‫‪ ٧-٤‬ﺣﺪود اﻟﺘﻨﺒﺆ ‪٦٨ .........................................................................................‬‬
‫‪ ١-٧-٤‬ﺗﻌﺮﻳﻒ ﻓﺘﺮة ﺗﻨﺒﺆ ﻟﻠﻘﻴﻤﺔ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪٦٨ .....................................................‬‬
‫‪ ٢-٧-٤‬ﻣﺜﺎل ‪٦٩ ........................................................................................‬‬
‫‪ -٥‬اﻟﻔﺼﻞ اﻟﺨﺎﻣﺲ‪ :‬ﺗﺼﻤﻴﻢ وﺑﻨﺎء ﻧﻈﺎم ﺗﻨﺒﺆ إﺣﺼﺎﺋﻲ ‪٧١ ............................................‬‬
‫‪ ١-٥‬ﺗﻌﻴﻴﻦ أو ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ‪٧١ .......................................................................‬‬
‫‪ ١-١-٥‬ﺗﺜﺒﻴﺖ اﻟﺘﺒﺎﻳﻦ ‪٧١ ..............................................................................‬‬
‫‪ ٢-١-٥‬إﺧﺘﻴﺎر درﺟﺔ اﻟﺘﻔﺮﻳﻖ ‪٧١ ................................................................ d‬‬

‫‪٥‬‬

‫‪ ٣-١-٥‬ﺗﺤﺪﻳﺪ ‪٧١ ............................................................................... p,q‬‬
‫‪ ٤-١-٥‬إﺿﺎﻓﺔ ﻣﻌﻠﻢ إﻧﺠﺮاف ‪٧١ ...................................................................‬‬
‫‪ ٢-٥‬ﺗﻘﺪﻳﺮ اﻟﻨﻤﻮذج ‪٧٢ ..............................................................................‬‬
‫‪ ١-٢-٥‬ﻃﺮﻳﻘﺔ اﻟﻌﺰوم ‪٧٢ ........................................................................‬‬
‫‪ ٢-٢-٥‬ﺗﻘﺪﻳﺮ اﻟﻌﺰوم ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ‪٧٣ .......................................................‬‬
‫‪ ١-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٣ .............................................................. AR(١‬‬
‫‪ ٢-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٤ ......................................................... MA(١‬‬
‫‪ ٣-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٤ .......................................................... AR(٢‬‬
‫‪ ٤-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٤ ......................................................... MA(٢‬‬
‫‪ ٥-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٥ ................................................ ARMA(١،١‬‬
‫‪ ٣-٢-٥‬ﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ‪٧٥ ..............................................‬‬
‫‪ ٤-٢-٥‬ﺗﻘﺪﻳﺮات اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ‪٧٦ .........................‬‬
‫‪ ١-٤-٢-٥‬ﻟﻨﻤﺎذج )‪٧٦ .........................................................AR(١‬‬
‫‪ ٢-٤-٢-٥‬ﻟﻨﻤﺎذج )‪٧٧ ....................................................... MA(١‬‬
‫ﺗﺸﺨﻴﺺ وإﺧﺘﺒﺎر اﻟﻨﻤﻮذج‪٧٨ .........................................................‬‬
‫‪٣-٥‬‬
‫‪ ١-٣-٥‬ﻓﺤﺺ اﻟﺒﻮاﻗﻲ ‪٧٨ ....................................................................‬‬
‫‪ ١-١-٣-٥‬إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ ﻟﻠﺒﻮاﻗﻲ ‪٧٩ ..................................................‬‬
‫‪ ٢-١-٣-٥‬إﺧﺘﺒﺎر اﻟﻌﺸﻮاﺋﻴﺔ ﻟﻠﺒﻮاﻗﻲ ‪٧٩ .................................................‬‬
‫‪ ٣-١-٣-٥‬إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ أو اﻹﺳﺘﻘﻼل ﻟﻠﺒﻮاﻗﻲ ‪٧٩ ...................................‬‬
‫‪ ٤-١-٣-٥‬إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ ‪٧٩ ....................................................‬‬
‫‪ ٢-٣-٥‬ﺑﻌﺾ اﻟﻤﻌﺎﻳﻴﺮ اﻻﺧﺮى ﻹﺧﺘﻴﺎر ﻧﻤﻮذج ﻣﻨﺎﺳﺐ ‪٧٩ .............................‬‬
‫‪ ١-٢-٣-٥‬إﺣﺼﺎﺋﻴﺔ آﻴﻮ ﻟﻠﺠﻨﻖ وﺑﻜﺲ ‪٧٩ ...........................................‬‬
‫‪ ٢-٢-٣-٥‬ﻣﻌﻴﺎر اﻹﻋﻼم اﻟﺬاﺗﻲ ‪٨٠ ......................................... AIC‬‬
‫‪ ٣-٣-٥‬أﻣﺜﻠﺔ وﺣﺎﻻت دراﺳﺔ ‪٨٠ ..........................................................‬‬
‫‪ -٦‬اﻟﻔﺼﻞ اﻟﺴﺎدس‪ :‬ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻤﻮﺳﻤﻴﺔ ‪١١٣ ......‬‬
‫‪ ١-٦‬دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ ‪١١٤ .....‬‬
‫‪ ١-١-٦‬ﻟﻨﻤﻮذج ‪١١٤ ......................................... SARMA(٠،١)(١،١)١٢‬‬
‫‪ ٢-٦‬دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ ‪١١٥ ..................................‬‬
‫‪١١٥............................................. SARIMA(٠,d,٠)(٠,D,١)s ١-٢-٦‬‬
‫‪١١٥ ........................................... SARIMA(٠,d,٠)(١,D,١)s ٢-٢-٦‬‬
‫‪١١٥ ............................................ SARIMA(٠,d,١)(٠,D,١)s ٣-٢-٦‬‬
‫‪١١٦ ........................................... SARIMA(٠,d,٠)(١,D,١)s ٤-٢-٦‬‬
‫‪١١٦ ........................................... SARIMA(٠,d,١)(١,D,٠)s ٥-٢-٦‬‬
‫‪١١٦ ............................................ SARIMA(٠,d,٢)(٠,D,١)s ٦-٢-٦‬‬
‫‪ ٣-٦‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻨﻤﻮذج اﻟﻤﻮﺳﻤﻲ اﻟﺘﻀﺎﻋﻔﻲ ‪١١٨ ....................‬‬
‫‪ ٤-٦‬أﻣﺜﻠﺔ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ ‪١١٩ .......................................‬‬
‫‪ ٥-٦‬إﺷﺘﻘﺎق دوال ﺗﻨﺒﺆ ﻟﺒﻌﺾ ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﻤﻮﺳﻤﻴﺔ اﻟﺘﻀﺎﻋﻔﻴﺔ ‪١٢٣ ..........‬‬
‫‪ ١-٥-٦‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻠﻨﻤﻮذج ‪١٢٣ .................... SARIMA(٠،٠،٠)(٠،١،١)١٢‬‬
‫‪ ٢-٥-٦‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻠﻨﻤﻮذج ‪١٢٤ .................... SARIMA(٠،١،١)(٠،١،١)١٢‬‬
‫‪ ٦-٦‬أﻣﺜﻠﺔ وﺣﺎﻻت دراﺳﺔ ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ ‪١٢٥ ..........................‬‬
‫اﻟﺠﺰء اﻟﻌﻤﻠﻲ‪:‬‬
‫‪ -٧‬اﻟﻔﺼﻞ اﻟﺴﺎﺑﻊ‪ :‬ورﻗﺔ ﺗﺪرﻳﺐ ﻋﻤﻠﻲ ﻋﻠﻰ اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ‬
‫اﻟﻤﺘﺤﺮك ‪١٤٢ ............................................................................................‬‬
‫‪٦‬‬

‫‪ -٨‬اﻟﻔﺼﻞ اﻟﺜﺎﻣﻦ‪ :‬ﻣﺜﺎل ﺗﺤﻠﻴﻞ اﻟﺒﻮاﻗﻲ وﻣﻌﻴﻴﺮ إﺧﺘﻴﺎر ﻧﻤﻮذج ﻣﻨﺎﺳﺐ ‪١٥٥ ..................‬‬
‫‪ -٩‬اﻟﻔﺼﻞ اﻟﺘﺎﺳﻊ‪ :‬ﺗﺤﻠﻴﻞ أو ﺗﻔﻜﻴﻚ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ إﻟﻰ ﻣﺮآﺒﺎت ‪١٦٢ .....................‬‬
‫‪ -١٠‬اﻟﺘﻤﻬﻴﺪ واﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪١٧٨ ...........................................‬‬
‫‪ ١-١٠‬اﻟﻮﺳﻴﻂ اﻟﺠﺎري ‪١٨١ .............................................................‬‬
‫‪ -١١‬اﻟﻔﺼﻞ اﻟﺤﺎدي ﻋﺸﺮ‪ :‬اﻟﺘﻤﻬﻴﺪ واﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﺘﻤﻬﻴﺪ اﻷﺳﻲ اﻟﺒﺴﻴﻂ ‪١٨٤ .............‬‬
‫‪ -١٢‬اﻟﻔﺼﻞ اﻟﺜﺎﻧﻲ ﻋﺸﺮ‪ :‬اﻟﺘﻤﻬﻴﺪ واﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﺘﻤﻬﻴﺪ اﻷﺳﻲ اﻟﻤﺰدوج ‪١٩٠ ...........‬‬
‫‪ ١-١٢‬ﻃﺮﻳﻘﺔ ﺑﺮاون ‪١٩٠ ........................................................................‬‬
‫‪ ٢-١٢‬ﻃﺮﻳﻘﺔ هﻮﻟﺖ ‪١٩٠ ........................................................................‬‬
‫‪ ٣-١٢‬أﻣﺜﻠﺔ ‪١٩١ ..................................................................................‬‬
‫‪ -١٣‬اﻟﻔﺼﻞ اﻟﺜﺎﻟﺚ ﻋﺸﺮ‪ :‬اﻟﺘﻤﻬﻴﺪ اﻷﺳﻲ اﻟﺜﻼﺛﻲ واﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﻟﻠﻤﺘﺴﻠﺴﻼت‬
‫اﻟﻤﻮﺳﻤﻴﺔ اﻟﻤﻨﺠﺮﻓﺔ ‪١٩٨ .............................................................................‬‬
‫‪ ١-١٣‬اﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ ‪١٩٩ ...............................................................‬‬
‫‪ ٢-١٣‬اﻟﻨﻤﻮذج اﻟﺘﻀﺎﻋﻔﻲ ‪٢٠١ ..............................................................‬‬
‫‪ ٣-١٣‬ﻣﺜﺎل ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ ‪٢٠٤ ...........................................................‬‬
‫‪ ٤-١٣‬ﻣﺜﺎل ﺁﺧﺮ ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ ‪٢١٠ .....................................................‬‬
‫اﻟﻤﺮاﺟﻊ ‪٢١٩ ...........................................................................................‬‬

‫‪٧‬‬

‫اﻟﻔﺼﻞ اﻷول‬
‫ﻣﻘﺪﻣﺔ وﺗﻌﺎرﻳﻒ‬
‫ﺗﻌﺮﻳﻒ ‪ :١‬اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ ‪Time Series‬‬

‫ﻣﺘﺘﺎﺑﻌﺔ ﻣﻦ اﻟﻘﻴﻢ اﻟﻤﺸﺎهﺪة ﻟﻈﺎهﺮة ﻋﺸﻮاﺋﻴﺔ ﻣﺮﺗﺒﺔ ﻣﻊ اﻟﺰﻣﻦ ) او ﻣﺮﺗﺒﺔ ﻣﻊ اﻟﻤﻜﺎن (‬

‫اﻣﺜﻠﺔ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ‪:‬‬
‫‪ -١‬ﺳﻌﺮ اﻗﻔﺎل ﺳﻬﻢ ﺑﻨﻚ اﻟﺮﻳﺎض ﻳﻮﻣﻴﺎ‪.‬‬
‫‪ -٢‬ﻋﺪد اﻟﻮﺣﺪات اﻟﻤﻄﻠﻮﺑﺔ اﺳﺒﻮﻋﻴﺎ ﻣﻦ اﻧﺘﺎج ﺳﻠﻌﺔ ﻣﻌﻴﻨﺔ‪.‬‬
‫‪ -٣‬ﺣﺠﻢ اﻟﻤﺒﻴﻌﺎت ﺷﻬﺮﻳﺎ ﻣﻦ ﺳﻠﻌﺔ ﻣﺎ‪.‬‬
‫‪ -٤‬ﺣﺠﻢ اﻹﻧﺘﺎج اﻟﻴﻮﻣﻲ ﻟﻠﻨﻔﻂ اﻟﺨﺎم ﺑﺎﻟﻤﻤﻠﻜﺔ‪.‬‬
‫واﻟﻐﺮض ﻣﻦ دراﺳﺔ وﺗﺤﻠﻴﻞ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ هﻮ‪:‬‬
‫‪ -١‬ﻓﻬﻢ وﻧﻤﺬﺟﺔ ﻋﺸﻮاﺋﻴﺔ اﻟﻈﺎهﺮة اﻟﻤﺸﺎهﺪة‪.‬‬
‫‪ -٢‬اﻟﺘﻨﺒﺆ ﻋﻦ اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻟﻠﻈﺎهﺮة اﻟﻌﺸﻮاﺋﻴﺔ‪.‬‬
‫‪ -٣‬اﻟﺘﺤﻜﻢ ﺑﺎﻟﻈﺎهﺮة اﻟﻌﺸﻮاﺋﻴﺔ إذا اﻣﻜﻦ ذﻟﻚ‪.‬‬
‫اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة وهﻲ ﻋﺒﺎرة ﻋﻦ اﻹﻧﺘﺎج اﻟﻴﻮﻣﻲ ﻟﻠﺤﻠﻴﺐ ﺑﺎﻟﺮﻃﻞ ﻟﺒﻘﺮة ﻣﺎ‬

‫‪850‬‬

‫‪750‬‬

‫‪C1‬‬
‫‪650‬‬

‫‪550‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫‪30‬‬

‫‪40‬‬

‫‪20‬‬

‫‪10‬‬

‫‪Index‬‬

‫اﻟﺨﻄﻮات اﻟﻤﺘﺨﺬة ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ‪:‬‬
‫إن إﻳﺠﺎد ﻧﻤﻮذج ﻣﻨﺎﺳﺐ ﺗﻨﻄﺒﻖ ﻋﻠﻴﺔ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة ﻳﻌﺘﺒﺮ ﻣﻦ اﻟﻤﻬﺎم اﻟﺼﻌﺒﺔ واﻟﺘﻲ ﺗﺤﺘﺎج‬
‫اﻟﻰ اﻟﻜﺜﻴﺮ ﻣﻦ اﻟﺒﺤﺚ واﻟﺨﺒﺮة‪ .‬ﺳﻮف ﻧﺴﺘﻌﺮض ﺑﻌﺾ اﻟﺨﻄﻮات اﻟﻌﺮﻳﻀﺔ ﻟﺒﻨﺎء ﻧﻤﻮذج رﻳﺎﺿﻲ‬
‫ﻟﻠﺘﻨﺒﺆ ﻋﻦ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺎ‪:‬‬
‫‪-١‬‬
‫ﺗﻌﻴﻴﻦ اﻟﻨﻤﻮذج أو ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ‪ :Model Identification‬وهﺬا ﻳﺘﻢ ﺑﺮﺳﻢ‬
‫اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ ﻓﻴﻤﺎ ﻳﺴﻤﻰ ‪ Time Plot‬ﺣﻴﺚ ﻳﻜﻮن اﻹﺣﺪاﺛﻲ اﻻﻓﻘﻲ هﻮ اﻟﺰﻣﻦ واﻟﺮأﺳﻲ ﺣﺠﻢ‬
‫اﻟﻈﺎهﺮة اﻟﻤﺸﺎهﺪة وﻣﻦ ﺛﻢ إﺧﺘﻴﺎر ﻧﻤﻮذج رﻳﺎﺿﻲ ﻣﻌﺘﻤﺪﻳﻦ ﻋﻠﻲ ﺑﻌﺾ اﻟﻤﻘﺎﻳﻴﺲ اﻹﺣﺼﺎﺋﻴﺔ اﻟﺘﻰ‬
‫ﺗﻤﻴﺰ ﻧﻤﻮذج ﻋﻦ ﺁﺧﺮ وﻋﻠﻰ اﻟﺨﺒﺮة اﻟﻤﺴﺘﻤﺪة ﻣﻦ اﻟﺪراﺳﺎت واﻻﺑﺤﺎث‪.‬‬

‫‪٨‬‬

‫‪-٢‬‬

‫‪-٣‬‬

‫‪-٤‬‬

‫‪-٥‬‬

‫ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج ‪ :Model Fitting‬ﺑﻌﺪ ﺗﺮﺷﻴﺢ ﻧﻤﻮذج او اآﺜﺮ آﻨﻤﻮذج ﻣﻨﺎﺳﺐ ﻟﻮﺻﻒ‬
‫اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺸﺎهﺪة ﻧﻘﻮم ﺑﺘﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ هﺬا اﻟﻨﻤﻮذج ﻣﻦ اﻟﺒﻴﺎﻧﺎت اﻟﻤﺸﺎهﺪة ﺑﺈﺳﺘﺨﺪام ﻃﺮق‬
‫اﻟﺘﻘﺪﻳﺮ اﻹﺣﺼﺎﺋﻲ اﻟﺨﺎﺻﺔ ﺑﺎﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ وهﺬا اﻟﻨﻤﻮذج اﻟﻤﺮﺷﺢ ﻳﺆﺧﺬ آﻨﻤﻮذج‬
‫اوﻟﻲ ﻗﺎﺑﻞ ﻟﻠﺘﻌﺪﻳﻞ ﻻﺣﻘﺎ‪.‬‬
‫ﺗﺸﺨﻴﺺ وإﺧﺘﺒﺎر اﻟﻨﻤﻮذج ‪ :Model Diagnostics‬إﺟﺮاء إﺧﺘﺒﺎرات ﺗﻔﺤﺼﻴﺔ ﻋﻠﻰ‬
‫أﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ ‪ Fitting Errors‬ﻟﻤﻌﺮﻓﺔ ﻣﺪى ﺗﻄﺎﺑﻖ اﻟﻤﺸﺎهﺪات ﻣﻊ اﻟﻘﻴﻢ اﻟﻤﺤﺴﻮﺑﺔ ﻣﻦ‬
‫اﻟﻨﻤﻮذج اﻟﻤﺮﺷﺢ وﻣﺪى ﺻﺤﺔ ﻓﺮﺿﻴﺎت اﻟﻨﻤﻮذج‪ .‬ﻓﻲ ﺣﺎﻟﺔ إﺟﺘﻴﺎز اﻟﻨﻤﻮذج اﻟﻤﺮﺷﺢ ﻟﻬﺬﻩ‬
‫اﻹﺧﺘﺒﺎرات ﻧﻘﻮم ﺑﺈﻋﺘﻤﺎدة ﻋﻠﻰ اﻧﻪ اﻟﻨﻤﻮذج اﻟﻨﻬﺎﺋﻲ وﻳﺴﺘﺨﺪم ﻟﺘﻮﻟﻴﺪ ﺗﻨﺒﺆات ﻟﻠﻘﻴﻢ‬
‫اﻟﻤﺴﺘﻘﺒﻠﻴﺔ وإﻻ ﻧﻌﻮد ﻟﻠﺨﻄﻮة اﻻوﻟﻰ ﻟﺘﻌﻴﻴﻦ ﻧﻤﻮذج ﺟﺪﻳﺪ‪.‬‬
‫ﺗﻮﻟﻴﺪ اﻟﺘﻨﺒﺆات ‪ :Forecast Generation‬ﻳﺴﺘﺨﺪم اﻟﻨﻤﻮذج اﻟﻨﻬﺎﺋﻲ ﻟﺘﻮﻟﻴﺪ ﺗﻨﺒﺆات ﻋﻦ‬
‫اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ وﻣﻦ ﺛﻢ ﺣﺴﺎب أﺧﻄﺎء اﻟﺘﻨﺒﺆ ‪ Forecast Errors‬آﻠﻤﺎ اﺳﺘﺠﺪت ﻗﻴﻢ‬
‫ﺟﺪﻳﺪة ﻣﺸﺎهﺪة ﻣﻦ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ وﻣﺮاﻗﺒﺔ هﺬﻩ اﻷﺧﻄﺎء ﻓﻰ ﻣﺎ ﻳﺴﻤﻰ ﺑﻤﺨﻄﻄﺎت‬
‫اﻟﻤﺮاﻗﺒﺔ ‪ Control Charts‬واﻟﺘﻲ ﺗﻮﺿﻊ ﻟﻠﻘﺒﻮل ﺑﻨﺴﺒﺔ ﺧﻄﺄ ﻣﻌﻴﻦ إذا ﺗﺠﺎوزﺗﺔ أﺧﻄﺎء‬
‫اﻟﺘﻨﺒﺆ ﻳﻌﺎد اﻟﻨﻈﺮ ﻓﻲ اﻟﻨﻤﻮذج وﺗﻌﺎد اﻟﺪورة ﻣﻦ ﺟﺪﻳﺪ ﺑﺘﺤﺪﻳﺪ ﻧﻤﻮذج ﻣﺮﺷﺢ ﺁﺧﺮ‪.‬‬
‫إﺳﺘﺨﺪام اﻟﺘﻨﺒﺆات ووﺿﻊ اﻟﻘﺮارات ‪:Implementation and Decision making‬‬
‫ﺗﻘﺪم اﻟﺘﻨﺒﺆات ﻓﻰ ﺗﻘﺮﻳﺮ ﻟﺼﺎﻧﻌﻲ اﻟﻘﺮار ﻟﻠﻨﻈﺮ ﻓﻲ إﺳﺘﺨﺪاﻣﻬﺎ ﺑﺎﻟﺸﻜﻞ اﻟﻤﻨﺎﺳﺐ‪.‬‬

‫ﺗﻌﺎرﻳﻒ وﻣﺒﺎدئ اوﻟﻴﺔ‪:‬‬
‫ﺳﻮف ﻧﺮﻣﺰ ﻟﻠﻈﺎهﺮة اﻟﻌﺸﻮاﺋﻴﺔ أو اﻟﻌﻤﻠﻴﺔ اﻟﻌﺸﻮاﺋﻴﺔ اﻟﺘﻲ ﺗﻮﻟﺪ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ ﺑﺎﻟﺮﻣﺰ‬
‫}‪ {L, Z −1 , Z 0 , Z1 , Z 2 ,L‬او اﺧﺘﺼﺎرا }}‪ {Z t , t ∈{L, −1, 0,1, 2,L‬او ﺑﺒﺴﺎﻃﺔ‬

‫وﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪة ﺑﺎﻟﺮﻣﺰ } ‪{z1 , z2 ,L, zn−1 , zn‬‬

‫} ‪{Z t‬‬

‫ﺗﻌﺮﻳﻒ‪ :٢‬اﻟﻘﻴﻢ ‪ z1 , z2 ,L , zn −1‬ﺗﺴﻤﻰ ﺑﺎﻟﻤﺎﺿﻰ او ﺗﺎرﻳﺦ اﻟﻈﺎهﺮة ‪History‬‬

‫واﻟﺘﺎرﻳﺦ ﻣﻬﻢ ﺟﺪا ﻓﻲ ﻋﻤﻠﻴﺔ اﻟﻨﻤﺬﺟﺔ‬

‫ﺗﻌﺮﻳﻒ ‪ : ٣‬اﻟﻘﻴﻤﺔ ‪ zn‬ﺗﺴﻤﻰ اﻟﺤﺎﺿﺮ او اﻵن‬
‫وهﻲ اﻟﻤﺸﺎهﺪة اﻷﺧﻴﺮة ‪.‬‬

‫ﺗﻌﺮﻳﻒ ‪ :٤‬أﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ ‪ et = zt − zˆt , t = 1, 2,..., n‬ﺣﻴﺚ ‪ zˆt‬هﻲ اﻟﻘﻴﻢ‬
‫اﻟﻤﻄﺒﻘﺔ ) اﻟﻘﻴﻢ اﻟﺘﻲ ﻧﺘﺤﺼﻞ ﻋﻠﻴﻬﺎ ﻣﻦ اﻟﻨﻤﻮذج( وﺗﺴﻤﻲ أﻳﻀﺎ اﻟﺮواﺳﺐ ‪Residuals‬‬
‫وﻳﻼﺣﻆ ان اﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ ﻧﺤﺼﻞ ﻋﻠﻴﻬﺎ دﻓﻌﺔ واﺣﺪة ﺑﻌﺪ ﺗﻘﺪﻳﺮ اﻟﻨﻤﻮذج‪.‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬ﺳﻮف ﻧﺮﻣﺰ ﻟﻠﻤﺸﺎهﺪات اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺑﺎﻟﺮﻣﻮز ‪ zn +1 , zn + 2 , zn +3 ,...‬او ﺑﺸﻜﻞ ﻋﺎم‬
‫‪ zn +l , l ≥ 0‬وﻧﺮﻣﺰ ﻟﺘﻨﺒﺆاﺗﻬﺎ ﺑﺎﻟﺮﻣﺰ ‪ zn (1) , zn ( 2 ) , zn ( 3) ,...‬او ﺑﺸﻜﻞ ﻋﺎم‬
‫‪zn ( l ) , l ≥ 0‬‬

‫‪٩‬‬

‫‪en ( l ) = zn +l − zn ( l ) , l ≥ 0‬‬

‫ﺗﻌﺮﻳﻒ ‪ :٥‬أﺧﻄﺎء اﻟﺘﻨﺒﺆ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬

‫وأﺧﻄﺎء اﻟﺘﻨﺒﺆ ﺗﻨﺘﺞ اﻟﻮاﺣﺪة ﺗﻠﻮ اﻻﺧﺮى آﻠﻤﺎ ﺗﻘﺪم اﻟﺰﻣﻦ وﺷﻮهﺪت اﻟﻘﻴﻢ اﻟﺤﻘﻴﻘﻴﺔ‬
‫ﺗﻌﺮﻳﻒ ‪ :٦‬ﻳﻘﺎل ان اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪة } ‪ {z1 , z2 ,L, zn−1 , zn‬ﻣﺴﺘﻘﺮة ‪Stationary‬‬

‫إذا ﺣﻘﻘﺖ اﻟﺸﺮوط اﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪1) E ( zt ) = constant = µ , ∀t‬‬
‫‪⎧constant = γ 0 , ∀t , ∀s, t = s‬‬
‫⎨ = ) ‪2) cov ( zt , zs‬‬
‫‪⎩ f ( s − t ) , ∀t , ∀ s , t ≠ s‬‬

‫اﻵن ﺳﻮف ﻧﻌﺮف ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﻬﻤﺔ ﺟﺪا ﻟﻜﻮﻧﻬﺎ ﺣﺠﺮة او ﻃﻮب اﻟﺒﻨﺎء ‪Building Blocks‬‬
‫ﻟﺠﻤﻴﻊ اﻟﻨﻤﺎذج اﻟﺘﻲ ﺳﻮف ﻧﺪرﺳﻬﺎ‬

‫ﺗﻌﺮﻳﻒ ‪ :٧‬ﻣﺘﺴﻠﺴﻠﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪ White Noise Series‬اوﻋﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬
‫‪ {at } White Noise Process‬هﻲ ﻋﺒﺎرة ﻋﻦ ﻣﺘﺘﺎﺑﻌﺔ ﻣﻦ اﻟﻤﺸﺎهﺪات اﻟﻌﺸﻮاﺋﻴﺔ ﻏﻴﺮ‬
‫اﻟﻤﺘﺮاﺑﻄﺔ ) واﺣﻴﺎﻧﺎ ﻧﻔﺘﺮض اﻧﻬﺎ ﻣﺘﺘﺎﺑﻌﺔ ﻣﻦ اﻟﻤﺘﻐﻴﺮات اﻟﻌﺸﻮاﺋﻴﺔ اﻟﺘﻲ ﺗﻜﻮن ﻣﺴﺘﻘﻠﺔ وﻟﻬﺎ‬
‫ﺗﻮزﻳﻌﺎت ﻣﺘﻄﺎﺑﻘﺔ )‪ ( Independent, Identically Distributed (IID‬ﺑﻤﺘﻮﺳﻂ ﺻﻔﺮي‬
‫وﺗﺒﺎﻳﻦ ﺛﺎﺑﺖ ‪ σ 2‬أي‪:‬‬

‫‪1) E ( at ) = 0, ∀t‬‬

‫وﻳﺮﻣﺰ ﻟﻬﺎ ﺑﺎﻟﺮﻣﺰ ) ‪WN ( 0, σ 2‬‬

‫‪⎧σ 2 , ∀t , ∀s, t = s‬‬
‫⎨ = ) ‪2) cov ( at , as‬‬
‫‪⎩ 0 , ∀t , ∀ s , t ≠ s‬‬
‫‪at‬‬

‫ﻣﺜﺎل‪ :١‬ﻣﺘﺴﻠﺴﻠﺔ اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ‪: Random Walk‬‬
‫ﺳﻮف ﻧﺒﻨﻲ ﻋﻤﻠﻴﺔ ﻋﺸﻮاﺋﻴﺔ } ‪ {Z t‬آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫‪Z1 = a1‬‬
‫‪Z 2 = a1 + a2‬‬
‫‪M‬‬

‫‪Z t = a1 + a2 +L + at‬‬
‫أو‬

‫‪Z t = Z t −1 + at‬‬
‫أي ﻟﻮ اﻋﺘﺒﺮﻧﺎ ان ‪ a j‬هﻮ ﺣﺠﻢ اﻟﺨﻄﻮة اﻟﺘﻲ ﺗﺆﺧﺬ اﻟﻲ اﻻﻣﺎم او اﻟﺨﻠﻒ ﻋﻨﺪ اﻟﺰﻣﻦ ‪ j‬ﻓﺎن ‪Z t‬‬

‫هﻲ ﻣﻮﻗﻊ ﻣﺎﺷﻲ ﻋﺸﻮاﺋﻲ ﻋﻨﺪ اﻟﺰﻣﻦ ‪t‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬هﺬﻩ اﻟﻌﻤﻠﻴﺔ او اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﻦ اﻟﻨﻤﺎذج اﻟﻬﺎﻣﺔ ﺟﺪا اﻟﺘﻲ ﺗﺼﻒ اﺳﻮاق اﻟﻤﺎل اﻟﻌﺎﻟﻤﻴﺔ‬
‫ﺗﻤﺮﻳﻦ‪ :‬اوﺟﺪ ) ‪ E ( Z t‬و ) ‪ cov ( Z t , Z s‬ﻟﺠﻤﻴﻊ ﻗﻴﻢ ‪ t , s‬وهﻞ اﻟﻌﻤﻠﻴﺔ ﻣﺴﺘﻘﺮة؟‬
‫‪١٠‬‬

‫ﺗﻌﺮﻳﻒ ‪ :٨‬داﻟﺔ اﻟﺘﻐﺎﻳﺮ اﻟﺬاﺗﻲ ‪ Autocovariance Function‬وﺗﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪= cov ( Z t , Z s ) , ∀t , ∀s‬‬

‫‪γ t ,s‬‬

‫‪= E ⎡⎣( Z t − µ )( Z s − µ ) ⎤⎦ , ∀t , ∀s‬‬

‫وإذا ﻋﺮﻓﻨﺎ اﻟﺘﺨﻠﻒ ‪ k‬ﻋﻠﻲ اﻧﻪ اﻟﻔﺘﺮة اﻟﺰﻣﻨﻴﺔ اﻟﺘﻲ ﺗﻔﺼﻞ ﺑﻴﻦ ‪ Z t‬وﺑﻴﻦ ‪ Z t −k‬أو ‪ Z t +k‬ﻓﺈن‬
‫داﻟﺔ اﻟﺘﻐﺎﻳﺮ اﻟﺬاﺗﻲ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬
‫‪γ k = cov ( Z t , Z t −k ) , k = 0, ±1, ±2,L‬‬
‫‪= E ⎡⎣( Z t − µ )( Z t −k − µ ) ⎤⎦ , k = 0, ±1, ±2,L‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﺘﻌﺮﻳﻒ اﻟﺜﺎﻧﻲ داﺋﻤﺎ‬

‫ﺗﻌﺮﻳﻒ ‪ :٩‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ )‪ Autocorrelation Function (ACF‬وﺗﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪γ‬‬
‫‪ρ k = k , k = 0, ±1, ±2,L‬‬
‫‪γ0‬‬
‫وﻟﻬﺎ اﻟﺨﻮاص اﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪1. ρ 0 = 1‬‬
‫‪2. ρ − k = ρ k‬‬

‫‪ρk ≤ 1‬‬

‫‪3.‬‬

‫ﻣﺜﺎل ‪ :٢‬ﺳﻮف ﻧﺸﺘﻖ اﻵن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻌﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬
‫داﻟﺔ اﻟﺘﻐﺎﻳﺮ اﻟﺬاﺗﻲ ﻟﻌﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء هﻲ‪:‬‬
‫‪2‬‬
‫‪⎧σ , k = 0‬‬
‫⎨ = ) ‪γ k = cov ( at , at −k‬‬
‫‪⎩ 0, k ≠0‬‬
‫وذﻟﻚ ﻣﻦ اﻟﺘﻌﺮﻳﻒ ‪ ٧‬وﻣﻨﻬﺎ ﻧﺠﺪ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‪:‬‬
‫‪⎧1 , k = 0‬‬
‫‪γ‬‬
‫⎨ = ‪ρk = k‬‬
‫‪γ 0 ⎩0 , k ≠ 0‬‬
‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪Autocorrelation function of White Noise‬‬
‫‪1.0‬‬

‫‪0.0‬‬
‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪5‬‬

‫‪4‬‬

‫‪Lag‬‬

‫‪١١‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫‪Autocorr‬‬

‫‪0.5‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١٠‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ‪Partial Autocorrelation Function‬‬
‫)‪(PACF‬‬
‫وﺗﻌﻄﻲ ﻣﻘﺪار اﻟﺘﺮاﺑﻂ ﺑﻴﻦ ‪ Z t‬و ‪ Z t −k‬ﺑﻌﺪ إزاﻟﺖ ﺗﺄﺛﻴﺮ اﻟﺘﺮاﺑﻂ اﻟﻨﺎﺗﺞ ﻣﻦ اﻟﻤﺘﻐﻴﺮات‬
‫‪ Z t −1 , Z t −2 ,..., Z t −k +1‬اﻟﻮاﻗﻌﺔ ﺑﻴﻨﻬﻤﺎ وﻳﺮﻣﺰ ﻟﻬﺎ ﻋﻨﺪ اﻟﺘﺨﻠﻒ ‪ k‬ﺑﺎﻟﺮﻣﺰ ‪ φkk‬وأﺣﺪ ﻃﺮق‬
‫ﺣﺴﺎﺑﻬﺎ ﺗﻘﻮم ﻋﻠﻲ ﺣﺴﺎب ﻣﻌﺎﻣﻞ اﻹﻧﺤﺪار اﻟﺠﺰﺋﻲ ‪ φkk‬ﻓﻲ اﻟﺘﻤﺜﻴﻞ‪:‬‬
‫‪Z t = φk 1Z t −1 + φk 2 Z t −2 + L + φkk Z t −k + at‬‬

‫ﺣﺴﺎب ‪: φ11‬‬

‫‪Z t = φ11Z t −1 + at‬‬

‫ﺑﻀﺮب ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﺑـ ‪ Z t −1‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬

‫) ‪E ( Z t −1Z t ) = φ11 E ( Z t −1Z t −1 ) + E ( Z t −1at‬‬

‫أي‬

‫‪γ 1 = φ11γ 0‬‬
‫ﺣﻴﺚ ‪ ) E ( Z t −1at ) = 0‬ﺑﺸﻜﻞ ﻋﺎم ‪ E ( Z t −k at ) = 0, k = 1, 2,...‬آﻤﺎ ﺳﻨﺒﻴﻦ ﻻﺣﻘﺎ (‬
‫وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻲ ‪ γ 0‬ﻧﺠﺪ‬
‫‪φ11 = ρ1‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١١‬ﺑﺸﻜﻞ ﻋﺎم ﺗﻌﺮف داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪k =0‬‬
‫‪k =1‬‬

‫‪k = 2,3,...‬‬

‫‪1,‬‬

‫‪ρ1 ,‬‬
‫‪ρ1‬‬
‫‪ρ2‬‬

‫‪ρ k −2‬‬
‫‪ρ k −3‬‬

‫‪M‬‬

‫‪M‬‬

‫‪ρk‬‬
‫‪,‬‬
‫‪ρ k −1‬‬
‫‪ρ k −2‬‬
‫‪M‬‬
‫‪1‬‬

‫ﺣﻴﺚ‬

‫‪ρ1‬‬
‫‪ρ k −2‬‬
‫‪ρ k −3‬‬
‫‪M‬‬

‫‪ρ1‬‬

‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬

‫‪ρ1‬‬

‫‪1‬‬

‫‪1‬‬
‫‪M‬‬

‫‪ρ1‬‬
‫‪M‬‬

‫‪ρ k −1 ρ k −2‬‬
‫‪1‬‬
‫‪ρ1‬‬
‫‪1‬‬
‫‪ρ1‬‬
‫‪M‬‬

‫‪M‬‬

‫‪ρ k −1 ρ k −2‬‬

‫⎧‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎨ = ‪φkk‬‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎪‬
‫⎩⎪‬

‫ﺗﺮﻣﺰ اﻟﻲ ﻣﺤﺪدة ﻣﺼﻔﻮﻓﺔ‬

‫اﻟﺘﻌﺮﻳﻒ اﻟﺴﺎﺑﻖ ﺻﻌﺐ اﻹﺳﺘﺨﺪام ﻟﻘﻴﻢ ‪ k‬اﻟﻜﺒﻴﺮة وﻟﻬﺬا ﺳﻮف ﻧﻌﻄﻲ ﺗﻌﺮﻳﻒ ﺁﺧﺮ ﻟﺤﺴﺎب داﻟﺔ‬
‫اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﻜﺮارﻳﺎ‪:‬‬
‫‪١٢‬‬

‫ﺗﻌﺮﻳﻒ ‪ ١١‬ب‪ :‬ﺗﺤﺴﺐ ‪ φkk‬ﺗﻜﺮارﻳﺎ ﻣﻦ اﻟﻌﻼﻗﺎت‬

‫‪φ00 = 1, by definition‬‬
‫‪φ11 = ρ1‬‬
‫‪k −1‬‬

‫‪, k = 2,3,...‬‬

‫‪ρ k − ∑φk −1, j ρ k − j‬‬
‫‪j =1‬‬
‫‪k −1‬‬

‫‪1 − ∑φk −1, j ρ j‬‬

‫= ‪φkk‬‬

‫‪j =1‬‬

‫ﺣﻴﺚ‬
‫‪φkj = φk −1, j − φkkφk −1,k −1 ,‬‬

‫‪j = 1, 2,..., k − 1‬‬
‫ﺣﺴﺎب ‪: φ22‬‬
‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ١١‬ب‪:‬‬

‫‪ρ 2 − φ11 ρ1 ρ 2 − ρ12‬‬
‫= ‪φ22‬‬
‫=‬
‫‪1 − φ11 ρ1‬‬
‫‪1 − ρ12‬‬

‫وذﻟﻚ ﻷن ‪. φ11 = ρ1‬‬

‫ﻣﺜﺎل ‪ :٣‬ﺳﻮف ﻧﺸﺘﻖ اﻵن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻌﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‪:‬‬
‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ١١‬ب‬
‫‪φ00 = 1, by definition‬‬
‫‪φ11 = ρ1 = 0‬‬
‫وذﻟﻚ ﻣﻦ ﻣﺜﺎل ‪ ١‬اﻟﺴﺎﺑﻖ‬
‫وﺑﺎﻟﺘﻌﻮﻳﺾ ﻓﻲ ﺗﻌﺮﻳﻒ ‪ ١١‬ب ﻋﻦ ‪ φkk‬ﻧﺠﺪ‬
‫‪φ22 = φ33 = L = 0‬‬
‫وهﻜﺬا‪:‬‬
‫‪⎧1, k = 0‬‬
‫⎨ = ‪φkk‬‬
‫‪⎩0, k ≠ 0‬‬
‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪Partial Autocorrelation function of White Noise‬‬
‫‪1.0‬‬

‫‪0.0‬‬

‫‪PACF‬‬

‫‪0.5‬‬

‫ﻣﻼﺣﻈﺔ‪ :‬ﻻﺣﻆ أن آﻞ ﻣﻦ داﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻌﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬
‫‪Lag‬‬
‫ﺗﺴﺎوي اﻟﺼﻔﺮ ﻣﻦ اﻟﺘﺨﻠﻒ اﻷول‪ .‬وهﺬﻩ ﺧﺎﺻﻴﺔ ﺟﻤﻴﻊ اﻟﻤﺘﻐﻴﺮات اﻟﻌﺸﻮاﺋﻴﺔ ﻏﻴﺮ اﻟﻤﺘﺮاﺑﻄﺔ او‬
‫اﻟﻤﺴﺘﻘﻠﺔ‪ .‬ﻹﺧﺘﺒﺎر ﻋﺪم اﻟﺘﺮاﺑﻂ ﺑﻴﻦ ﻗﻴﻢ ﻣﺸﺎهﺪة ﻟﻤﺘﻐﻴﺮ ﻋﺸﻮاﺋﻲ ﺗﺴﺘﺨﺪم داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﺬﻟﻚ‪.‬‬
‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪5‬‬

‫‪4‬‬

‫‪١٣‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫ﺗﻌﺮﻳﻒ ‪ : ١٢‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻌﻴﻨﺔ ‪Sample Autocorrelation Function SACF‬‬
‫ﻟﻤﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة ‪ z1 , z2 ,L , zn −1 , zn‬وﻳﺮﻣﺰ ﻟﻬﺎ ﺑﺎﻟﺮﻣﺰ ‪ rk , k = 0,1, 2,...‬وﺗﻌﻄﻰ‬

‫ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬

‫‪, k = 0,1, 2, ...‬‬

‫) ‪− z )( zt +k − z‬‬
‫‪2‬‬

‫)‪−z‬‬

‫‪n −k‬‬

‫‪t‬‬

‫‪∑( z‬‬
‫‪t =1‬‬

‫‪n‬‬

‫‪t‬‬

‫‪∑( z‬‬

‫= ‪rk‬‬

‫‪t =1‬‬

‫‪1 n‬‬
‫ﺣﻴﺚ ‪z = ∑ zt‬‬
‫‪n t =1‬‬
‫وهﻲ ُﻣﻘ ﱢﺪر ‪ Estimator‬ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ أي ‪ ρˆ k = rk , k = 0,1, 2,...‬وﺑﻤﺎ اﻧﻬﺎ ُﻣﻘ ﱢﺪر‬
‫ﻓﻬﻲ إذا ﺗﺘﻐﻴﺮ ﻋﺸﻮاﺋﻴﺎ ﻣﻦ ﻋﻴﻨﺔ ﻻﺧﺮى وﻟﻬﺬا ﻓﺈن ﻟﻬﺎ اﻟﺨﻮاص اﻟﻌﻴﻨﻴﺔ اﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ -١‬إذا آﺎﻧﺖ ‪ ρ k = 0, k > q‬ﻓﺈن‬
‫‪q‬‬
‫⎛‪1‬‬
‫⎞‬
‫‪V ( rk ) ≅ ⎜ 1 + 2∑ ρ k2 ⎟ , k > q‬‬
‫⎝‪n‬‬
‫‪k =1‬‬
‫⎠‬
‫‪1‬‬
‫وﻓﻲ اﻟﺤﺎﻟﺔ اﻟﺨﺎﺻﺔ ﻋﻨﺪﻣﺎ ‪ ρ k = 0, k > 0‬ﻓﺈن ‪V ( rk ) ≅ , k > 0‬‬
‫‪n‬‬
‫‪ -٢‬ﻟﻘﻴﻢ ‪ n‬اﻟﻜﺒﻴﺮة و ‪ ρ k = 0‬ﻓﺈن ‪ rk‬ﻳﻜﻮن ﻟﻬﺎ ﺗﻘﺮﻳﺒﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ وﺑﺎﻟﺘﺎﻟﻲ ﻧﺴﺘﻄﻴﻊ اﻟﻘﻴﺎم‬
‫ﺑﺎﻻﺧﺘﺒﺎر اﻟﺘﺎﻟﻲ‪:‬‬

‫‪H 0 : ρk = 0‬‬
‫‪H1 : ρ k ≠ 0‬‬
‫وذﻟﻚ ﺑﺈﺳﺘﺨﺪام اﻹﺣﺼﺎﺋﺔ‪:‬‬
‫‪= n rk‬‬

‫‪rk‬‬
‫‪− 12‬‬

‫‪n‬‬

‫وذﻟﻚ ﻋﻨﺪ ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬وﺗﺮﻓﺾ ‪ H 0‬إذا آﺎﻧﺖ ‪n rk > 1.96‬‬
‫‪ -٣‬ﺗﺤﺖ اﻟﻔﺮﺿﻴﺔ ‪ H 0 : ρ k = 0, ∀k‬ﻓﺈن ‪corr ( rk , rk − s ) ≅ 0, s ≠ 0‬‬
‫‪ُ -٤‬ﺗﻘ ﱠﺪر اﻟﺘﺒﺎﻳﻨﺎت ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻌﻴﻨﺔ آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫⎛‪1‬‬
‫⎞‬
‫‪Vˆ ( rk ) ≅ ⎜ 1 + 2∑ rk2 ⎟ , k > q‬‬
‫⎝‪n‬‬
‫‪k =1‬‬
‫⎠‬
‫ﺗﻌﺮﻳﻒ ‪ :١٣‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ ‪Sample Partial Autocorrelation‬‬
‫‪ Function SPACF‬ﻟﻤﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة ‪ z1 , z2 ,L , zn −1 , zn‬وﻳﺮﻣﺰ ﻟﻬﺎ ﺑﺎﻟﺮﻣﺰ‬
‫‪ rkk , k = 0,1, 2,...‬ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬
‫‪q‬‬

‫‪١٤‬‬

‫‪k =0‬‬
‫‪k =1‬‬
‫‪rk −2 r1‬‬
‫‪rk −3 r2‬‬
‫‪M‬‬
‫‪M‬‬
‫‪r1 rk‬‬
‫‪,‬‬
‫‪rk −2 rk −1‬‬
‫‪rk −3 rk −2‬‬
‫‪M‬‬
‫‪M‬‬
‫‪r1‬‬
‫‪1‬‬

‫‪k = 2,3,...‬‬

‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬
‫‪L‬‬

‫‪⎧ 1,‬‬
‫‪⎪ r,‬‬
‫‪⎪ 1‬‬
‫‪⎪ 1‬‬
‫‪r1‬‬
‫⎪‬
‫‪1‬‬
‫‪⎪ r1‬‬
‫‪⎪ M‬‬
‫‪M‬‬
‫⎪‬
‫‪rkk = ⎨ rk −1 rk −2‬‬
‫‪⎪ 1‬‬
‫‪r1‬‬
‫⎪‬
‫‪1‬‬
‫‪⎪ r1‬‬
‫‪⎪ M‬‬
‫‪M‬‬
‫⎪‬
‫‪⎪ rk −1 rk −2‬‬
‫⎪‬
‫⎩⎪‬

‫و ﻟﺤﺴﺎب داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ ﺗﻜﺮارﻳﺎ‪:‬‬

‫ﺗﻌﺮﻳﻒ ‪ ١٣‬ب‪ :‬ﺗﺤﺴﺐ ‪ rkk‬ﺗﻜﺮارﻳﺎ ﻣﻦ اﻟﻌﻼﻗﺎت‬

‫‪r00 = 1, by definition‬‬
‫‪r11 = r1‬‬
‫‪k −1‬‬

‫‪, k = 2,3, ...‬‬

‫‪rk − ∑ rk −1, j rk − j‬‬
‫‪j =1‬‬
‫‪k −1‬‬

‫‪1 − ∑ rk −1, j rj‬‬

‫= ‪rkk‬‬

‫‪j =1‬‬

‫‪j = 1, 2,..., k − 1‬‬

‫ﺣﻴﺚ‬

‫‪rkj = rk −1, j − rkk rk −1,k −1 ,‬‬

‫وهﻲ اﻳﻀﺎ ﻣﻘﺪﱠر ‪ Estimator‬ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ أي‬
‫‪ φˆkk = rkk , k = 0,1, 2,...‬وﺑﻤﺎ اﻧﻬﺎ ُﻣﻘ ﱢﺪر ﻓﻬﻲ إذا ﺗﺘﻐﻴﺮ ﻋﺸﻮاﺋﻴﺎ ﻣﻦ ﻋﻴﻨﺔ ﻻﺧﺮى وﻟﻬﺬا ﻓﺈن‬
‫ﻟﻬﺎ اﻟﺨﻮاص اﻟﻌﻴﻨﻴﺔ اﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪1‬‬
‫‪V ( rkk ) ≅ , k > 0 -١‬‬
‫‪n‬‬
‫‪ -٢‬ﻟﻘﻴﻢ ‪ n‬اﻟﻜﺒﻴﺮة ﻓﺈن ‪ rkk‬ﻳﻜﻮن ﻟﻬﺎ ﺗﻘﺮﻳﺒﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ وﺑﺎﻟﺘﺎﻟﻲ ﻧﺴﺘﻄﻴﻊ اﻟﻘﻴﺎم ﺑﺎﻻﺧﺘﺒﺎر‬
‫اﻟﺘﺎﻟﻲ‪:‬‬

‫‪H 0 : φkk = 0‬‬
‫‪H1 : φkk ≠ 0‬‬
‫وذﻟﻚ ﺑﺈﺳﺘﺨﺪام اﻹﺣﺼﺎﺋﺔ‪:‬‬
‫‪١٥‬‬

‫‪= n rkk‬‬

‫‪rkk‬‬
‫‪− 12‬‬

‫‪n‬‬

‫وذﻟﻚ ﻋﻨﺪ ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬وﺗﺮﻓﺾ ‪ H 0‬إذا آﺎﻧﺖ ‪n rkk > 1.96‬‬

‫‪ -٣‬ﺗﺤﺖ اﻟﻔﺮﺿﻴﺔ ‪ H 0 : φkk = 0, ∀k‬ﻓﺈن ‪corr (φkk ,φk − s ,k −s ) ≅ 0, s ≠ 0‬‬

‫‪ُ -٤‬ﺗﻘ ّﺪر اﻟﺘﺒﺎﻳﻨﺎت ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻌﻴﻨﺔ آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪1‬‬
‫‪Vˆ ( rkk ) ≅ , k > 0‬‬
‫‪n‬‬
‫ﻣﺜﺎل ‪ :٤‬اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ ﺗﻤﺜﻞ اﻟﻄﻠﺐ ﻋﻠﻲ ﻣﻨﺘﺞ ﻣﻌﻴﻦ ﻳﻮﻣﻴﺎ‪:‬‬
‫‪158 222 248 216 226 239 206 178 169‬‬
‫أﺣﺴﺐ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ وارﺳﻤﻬﻤﺎ‪:‬‬
‫‪1‬‬
‫‪1‬‬
‫اوﻻ‪ :‬ﻧﺤﺴﺐ اﻟﻤﺘﻮﺳﻂ ‪zt = (158 + 222 + L + 169 ) = 206.89‬‬
‫∑‬
‫‪9‬‬
‫‪n t =1‬‬
‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﺤﺴﺐ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻣﻦ اﻟﻌﻼﻗﺔ‬
‫‪n‬‬

‫‪, k = 0,1,2,...‬‬

‫=‪z‬‬
‫‪n −k‬‬

‫) ‪− z )( zt +k − z‬‬
‫‪2‬‬

‫)‪−z‬‬

‫‪t‬‬

‫‪∑( z‬‬

‫‪n‬‬

‫‪t‬‬

‫‪∑( z‬‬

‫‪t =1‬‬

‫= ‪rk‬‬

‫‪t =1‬‬

‫) ‪(158 × 222 + 222 × 248 + L + 178 × 169‬‬
‫‪= 0.265116‬‬
‫) ‪(158 × 158 + 222 × 222 + L + 169 × 169‬‬
‫) ‪(158 × 248 + 222 × 216 + L + 206 × 169‬‬
‫= ‪r2‬‬
‫‪= -0.212‬‬
‫) ‪(158 × 158 + 222 × 222 + L + 169 × 169‬‬
‫) ‪(158 × 216 + 222 × 226 + L + 239 × 169‬‬
‫= ‪r3‬‬
‫‪= −0.076‬‬
‫) ‪(158 × 158 + 222 × 222 + L + 169 × 169‬‬
‫= ‪r1‬‬

‫وهﻜﺬا ‪r8 = 0.230, r7 = 0.104, r6 = −0.242, r5 = −0.387, r4 = −0.183‬‬
‫ﺛﺎﻟﺜﺎ‪ :‬ﻧﺤﺴﺐ اﻟﺘﺒﺎﻳﻨﺎت ﻣﻦ‬
‫‪q‬‬
‫⎛‪1‬‬
‫⎞‬
‫‪Vˆ ( rk ) ≅ ⎜ 1 + 2∑ rk2 ⎟ , k > q‬‬
‫⎝‪n‬‬
‫‪k =1‬‬
‫⎠‬
‫‪1‬‬
‫≅ ) ‪Vˆ ( r1‬‬
‫‪9‬‬
‫‪1‬‬
‫‪1‬‬
‫‪2‬‬
‫‪Vˆ ( r2 ) ≅ (1 + 2 r12 ) = 1 + 2 ( 0.265) = 0.1267‬‬
‫‪n‬‬
‫‪9‬‬
‫‪1‬‬
‫‪1‬‬
‫‪2‬‬
‫‪2‬‬
‫‪Vˆ ( r3 ) ≅ 1 + 2 ( r12 + r22 ) = 1 + 2 ( 0.265) + ( −0.212 ) = 0.1367‬‬
‫‪n‬‬
‫‪9‬‬

‫)‬

‫))‬

‫(‬

‫(‬

‫‪١٦‬‬

‫( )‬

‫(‬

‫‪Vˆ ( r4 ) ≅ 0.138 Vˆ ( r5 ) ≅ 0.1454 Vˆ ( r6 ) ≅ 0.1787‬‬
‫اﻟﺦ…‬
‫راﺑﻌﺎ‪ :‬ﻧﺤﺴﺐ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ‪:‬‬

‫‪r00 = 1, by definition‬‬
‫‪r11 = r1 = 0.265‬‬

‫ﺛﻢ ﻧﺤﺴﺐ ﺑﺎﻗﻲ اﻟﺘﺮاﺑﻄﺎت ﻣﻦ اﻟﻌﻼﻗﺎت اﻟﺘﻜﺮارﻳﺔ‬
‫‪k −1‬‬

‫‪, k = 2,3, ...‬‬

‫‪rk − ∑ rk −1, j rk − j‬‬
‫‪j =1‬‬
‫‪k −1‬‬

‫‪1 − ∑ rk −1, j rj‬‬

‫= ‪rkk‬‬

‫‪j =1‬‬

‫ﺣﻴﺚ‬

‫‪j = 1, 2,..., k − 1‬‬

‫‪rkj = rk −1, j − rkk rk −1,k −1 ,‬‬
‫‪1‬‬

‫‪r2 − r11r1 ( −0.212 ) − ( 0.265)( 0.265) −0.282225‬‬
‫=‬
‫=‬
‫‪1 − r11r1‬‬
‫)‪1 − ( 0.265)( 0.265‬‬
‫‪0.929775‬‬

‫=‬

‫‪r2 − ∑ r1, j r2− j‬‬
‫‪j =1‬‬
‫‪1‬‬

‫‪1 − ∑ r1, j rj‬‬

‫= ‪r22‬‬

‫‪j =1‬‬

‫‪= −0.30354‬‬
‫ﻟﺤﺴﺎب ‪ r33‬ﻧﺤﺘﺎج اﻟﻰ ‪ r21‬وﺗﺤﺴﺐ ﻣﻦ‬
‫‪r21 = r11 − r22 r11 = 0.265 − ( −0.303)( 0.265) = 0.345295‬‬

‫‪2‬‬

‫) ‪r3 − ( r21r2 + r22 r1‬‬
‫) ‪1 − ( r21r1 + r22 r2‬‬

‫=‬

‫‪r3 − ∑ r2, j r3− j‬‬
‫‪j =1‬‬
‫‪k −1‬‬

‫‪1 − ∑ r2, j rj‬‬

‫= ‪r33‬‬

‫‪j=1‬‬

‫) )‪( −0.076 ) − ( ( 0.345)( −0.212 ) + ( −0.303)( 0.265‬‬
‫) ) ‪1 − ( ( 0.345)( 0.265) + ( −0.303)( −0.212‬‬

‫=‬

‫‪= 0.092‬‬

‫وهﻜﺬا ﻧﺤﺴﺐ ﺑﺎﻗﻲ اﻟﺘﺮاﺑﻄﺎت اﻟﺠﺰﺋﻴﺔ ﻟﻠﻌﻴﻨﺔ‬
‫‪r88 = 0.042, r77 = 0.013, r66 = −0.207, r55 = −0.294, r44 = −0.298‬‬
‫وﻟﻬﺎ ﺟﻤﻴﻌﺎ اﻟﺘﺒﺎﻳﻨﺎت ﺗﺴﺎوي ﺗﻘﺮﻳﺒﺎ ‪1 = 0.1111‬‬
‫‪9‬‬
‫ﺧﺎﻣﺴﺎ‪ :‬رﺳﻢ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ‬
‫‪Autocorrelation Function for Demand‬‬

‫‪6‬‬

‫‪١٧‬‬

‫‪3‬‬

‫‪4‬‬

‫‪2‬‬

‫‪Corr‬‬

‫‪1‬‬

‫‪Lag‬‬

‫‪LBQ‬‬

‫‪T‬‬

‫‪Corr‬‬

‫‪Lag‬‬

‫‪LBQ‬‬

‫‪T‬‬

‫‪13.66‬‬

‫‪0.52‬‬

‫‪0.23‬‬

‫‪8‬‬

‫‪0.87‬‬
‫‪1.50‬‬

‫‪1 0.27 0.80‬‬
‫‪2 -0.21 -0.59‬‬

‫‪Autocorrelation‬‬

‫‪8‬‬

‫‪7‬‬

‫‪5‬‬

‫‪1.0‬‬
‫‪0.8‬‬
‫‪0.6‬‬
‫‪0.4‬‬
‫‪0.2‬‬
‫‪0.0‬‬
‫‪-0.2‬‬
‫‪-0.4‬‬
‫‪-0.6‬‬
‫‪-0.8‬‬
‫‪-1.0‬‬

Partial Autocorrelation

Partial Autocorrelation Function for Demand
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

1

2

3

4

5

6

Lag PAC

T

Lag PAC

T

0.27
-0.30
0.09
-0.30
-0.29
-0.21
0.01

0.80
-0.91
0.27
-0.89
-0.88
-0.62
0.04

8 0.04

0.13

1
2
3
4
5
6
7

١٨

7

8

‫اﻟﻔﺼﻞ اﻟﺜﺎﻧﻲ‬
‫ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪Autoregressive-Moving‬‬
‫‪ Average Models‬وإﺳﺘﺨﺪاﻣﺎﺗﻬﺎ ﻓﻲ اﻟﺘﻨﺒﺆ‪:‬‬
‫هﻨﺎك ﻋﺎﺋﻠﺔ آﺒﻴﺮة ﻣﻦ اﻟﻨﻤﺎذج اﻟﺘﻲ ﻳﻄﻠﻖ ﻋﻠﻴﻬﺎ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‬
‫‪ Autoregressive-Moving Average Models‬واﻟﺘﻲ اﺛﺒﺘﺖ اﻷﺑﺤﺎث اﻟﻜﺜﻴﺮة ﻓﻲ ﻣﺨﺘﻠﻒ‬
‫اﻟﻤﻴﺎدﻳﻦ اﻟﺘﻄﺒﻴﻘﻴﺔ ﻋﻠﻲ ﺗﻔﻮﻗﻬﺎ اﻟﻬﺎﺋﻞ ﻋﻠﻲ اﻟﻄﺮق اﻟﺘﻘﻠﻴﺪﻳﺔ ﻓﻲ اﻟﺘﻨﺒﺆ‪.‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١٤‬ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ) ‪ ( p, q‬وﻳﺮﻣﺰ ﻟﻪ‬
‫) ‪ ARMA ( p, q‬ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة } ‪ {z1 , z2 ,K, zn−1 , zn‬ﻳﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬
‫‪zt = δ + φ1 zt −1 + φ2 zt −2 + L + φ p zt − p + at − θ1at −1 − θ 2 at −2 − L − θ q at −q‬‬

‫ﺣﻴﺚ ) ‪ at WN ( 0,σ 2‬ﻣﺘﺴﻠﺴﻠﺔ ﺿﺠﺔ ﺑﻴﻀﺎء و ∞ < ‪ −∞ < δ‬ﻣﻌﻠﻢ ﺛﺎﺑﺖ ﻳﻤﺜﻞ اﻟﻤﺴﺘﻮي‬
‫و ‪ φ1 ,φ2 ,K,φ p‬هﻲ ﻣﻌﺎﻟﻢ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ‪ Autoregressive Parameters‬و‬
‫‪ θ1 ,θ 2 ,K,θ q‬هﻲ ﻣﻌﺎﻟﻢ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪Moving Average Operators‬‬

‫ﺳﻮف ﻧﺴﺘﻌﻴﻦ ﺑﺠﺒﺮ اﻟﻌﻤﺎل ‪ Operators Algebra‬ﻟﺘﺒﺴﻴﻂ هﺬﻩ اﻟﻨﻤﺎذج ﻟﻜﻲ ﻳﺴﻬﻞ اﻟﺘﻌﺎﻣﻞ ﻣﻌﻬﺎ‬

‫ﺗﻌﺮﻳﻒ ‪ :١٥‬ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻟﺨﻠﻔﻲ ‪ Backshift Operator‬وﻳﺮﻣﺰ ﻟﻪ ‪ B‬وﻟﻪ اﻟﺨﻮاص‬
‫اﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪1 − Bzt = zt −1‬‬
‫‪2 − B m zt = B m−1 ( Bzt ) = B m−2 ( B ( Bzt ) ) = L = zt −m‬‬

‫‪3 − Bc = c, c is a constant‬‬
‫ﺑﺎﻹﺿﺎﻓﺔ اﻟﻲ ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻟﺨﻠﻔﻲ ﺗﻮﺟﺪ ﻋﻤﺎل اﺧﺮي ﻧﺤﺘﺎج اﻟﻴﻬﺎ ﻻﺣﻘﺎ هﻲ‪:‬‬

‫ﺗﻌﺮﻳﻒ ‪ ١٥‬ب‪:‬‬
‫‪ -١‬ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻷﻣﺎﻣﻲ ‪ Forewardshift Operator‬وﻳﺮﻣﺰ ﻟﻪ ‪ F‬وﻳﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪F = B −1‬‬
‫‪ -٢‬ﻋﺎﻣﻞ اﻟﺘﻔﺮﻳﻖ ‪ Difference Operator‬وﻳﺮﻣﺰ ﻟﻪ ∇ وﻳﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫) ‪∇ = (1 − B‬‬
‫‪ -٣‬ﻋﺎﻣﻞ اﻟﺘﺠﻤﻴﻊ ‪ Sum Operator‬وﻳﺮﻣﺰ ﻟﻪ ‪ S‬وﻳﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪−1‬‬
‫) ‪S = ∇ −1 = (1 − B‬‬
‫اﻵن ﻧﻌﻮد اﻟﻲ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ) ‪ ( p, q‬وﻧﻜﺘﺒﻪ ﻋﻠﻲ‬
‫اﻟﺸﻜﻞ‪:‬‬

‫‪١٩‬‬

zt − φ1 zt −1 − φ2 zt −2 − L − φ p zt − p = δ + at − θ1at −1 − θ 2 at −2 − L − θ q at −q
zt − φ1 Bzt − φ2 B 2 zt − L − φ p B p zt = δ + at − θ1 Bat − θ 2 B 2 at − L − θ q B q at

(1 − φ B − φ B
1

2

2

− L − φ p B p ) zt = δ + (1 − θ1 B − θ 2 B 2 − L − θ q B q ) at

‫أو‬

φ p ( B ) zt = δ + θ q ( B ) at

Autoregressive ‫ هﻮ ﻋﺎﻣﻞ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‬φ p ( B ) = 1 − φ1B − φ2 B 2 − L − φ p B p ‫ﺣﻴﺚ‬
‫ هﻮ ﻋﺎﻣﻞ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‬θ q ( B ) = 1 − θ1 B − θ 2 B 2 − L − θ q B q ‫ و‬Operator
Moving Average Operator

:‫أﻣﺜﻠﺔ‬
‫ وﻳﻜﺘﺐ‬ARMA ( 0,0 ) ‫ وﻳﺮﻣﺰ ﻟﻪ‬Constant Mean Model ‫ ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﺜﺎﺑﺖ‬-١
:‫ﻋﻠﻲ اﻟﺸﻜﻞ‬
φ0 ( B ) zt = δ + θ 0 ( B ) at
‫او‬
(1) zt = δ + (1) at
WN ( 0,σ 2 )

z t = δ + a t , at

:‫ وهﻮ ﻋﻠﻲ اﻟﺸﻜﻞ‬ARMA (1,0 ) ≡ AR (1) ‫ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻻوﻟﻲ‬-٢
φ1 ( B ) zt = δ + θ 0 ( B ) at

(1 − φ1B ) zt = δ + at

zt = δ + φ1 zt −1 + at , at

WN ( 0,σ 2 )

:‫ وهﻮ ﻋﻠﻲ اﻟﺸﻜﻞ‬ARMA ( 0,1) ≡ MA (1) ‫ ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻻوﻟﻲ‬-٣
φ0 ( B ) zt = δ + θ1 ( B ) at
zt = δ + (1 − θ1 B ) at

zt = δ + at − θ1at −1 , at

WN ( 0,σ 2 )

:‫ وهﻮ ﻋﻠﻲ اﻟﺸﻜﻞ‬ARMA ( 2,0 ) ≡ AR ( 2 ) ‫ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ‬-٤
φ 2 ( B ) z t = δ + θ 0 ( B ) at

(1 − φ B − φ B ) z
2

1

2

t

= δ + at

zt = δ + φ1 zt −1 + φ2 zt −2 + at , at

WN ( 0,σ 2 )

:‫ وﻧﻜﺘﺒﻪ ﻋﻠﻲ اﻟﺸﻜﻞ‬ARMA (1,1) (١‫و‬١) ‫ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ‬-٥

φ1 ( B ) zt = δ + θ1 ( B ) at
(1 − φ1B ) zt = δ + (1 − θ1B ) at

zt = δ + φ1 zt −1 + at − θ1at −1 , at

WN ( 0,σ 2 )

٢٠

‫ﺧﺼﺎﺋﺺ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪:‬‬
‫ﺳﻮف ﻧﺪرس اﻟﺨﺼﺎﺋﺺ اﻹﺣﺼﺎﺋﻴﺔ اﻟﺘﻲ ﺗﻤﻴﺰ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك وﻣﻌﺮﻓﺔ آﻴﻔﻴﺔ اﻟﺘﻌﺮف‬
‫ﻋﻠﻲ اﺣﺪ هﺬﻩ اﻟﻨﻤﺎذج ﻣﻦ ﻋﻴﻨﺔ ﻣﺸﺎهﺪة وذﻟﻚ ﻟﺘﻌﻴﻴﻦ او ﺗﺤﺪﻳﺪ ﻧﻤﻮذج ﻣﻨﺎﺳﺐ ﻳﺼﻒ اﻟﻤﺸﺎهﺪات‪.‬‬

‫أوﻻ‪ :‬ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﺜﺎﺑﺖ)‪:ARMA(٠،٠‬‬
‫وﻳﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‬

‫‪φ0 ( B ) zt = δ + θ 0 ( B ) at‬‬

‫او‬

‫) ‪WN ( 0, σ 2‬‬

‫‪z t = δ + a t , at‬‬

‫ﺳﻮف ﻧﺸﺘﻖ اﻟﺨﻮاص اﻹﺣﺼﺎﺋﻴﺔ ﻟﻬﺬا اﻟﻨﻤﻮذج وذﻟﻚ ﺑﺈﻳﺠﺎد اﻟﺘﻮﻗﻊ )اﻟﻤﺘﻮﺳﻂ( وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ‬
‫اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫) ‪E ( z t ) = δ + E ( at‬‬
‫وذﻟﻚ ﻷن ) ‪WN ( 0, σ 2‬‬

‫‪=δ‬‬
‫‪at‬‬

‫ﺳﻮف ﻧﺮﻣﺰ ﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺴﻠﺴﻠﺔ ) ‪ E ( zt‬ﺑﺎﻟﺮﻣﺰ ‪ µ‬أي ) ‪ µ = E ( zt‬وﺑﺎﻟﺘﺎﻟﻲ ﻳﻜﻮن ‪δ = µ‬‬

‫وﻳﻜﺘﺐ اﻟﻨﻤﻮذج‪:‬‬

‫‪zt − µ = at‬‬
‫ﻹﺷﺘﻘﺎق داﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻧﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻓﻲ‬
‫‪ zt −k − µ‬وﻧﺄﺧﺬ اﻟﺘﻮﻗﻊ أي‬

‫⎦⎤ ‪E ⎡⎣( zt −k − µ )( zt − µ ) ⎤⎦ = E ⎡⎣( zt −k − µ ) at‬‬

‫وﻟﻜﻦ ‪ E ⎡⎣( zt −k − µ )( zt − µ ) ⎤⎦ = γ k‬ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٨‬إذا‬
‫‪γ k = E ⎡⎣( zt −k − µ ) at ⎤⎦ , k = 0, ±1, ±2,L‬‬
‫وﻧﺤﻞ هﺬﻩ اﻟﻌﻼﻗﺔ ﺗﻜﺮارﻳﺎ‪:‬‬
‫⎦⎤ ‪k = 0 : γ 0 = E ⎡⎣( zt − µ ) at‬‬
‫ﻹﻳﺠﺎد اﻟﻄﺮف اﻷﻳﻤﻦ ﻧﻀﺮب ﻃﺮﻓﻲ ‪ zt − µ = at‬ﻓﻲ ‪ at‬وﻧﺄﺧﺬ اﻟﺘﻮﻗﻊ أي‬
‫وذﻟﻚ ﻷن ) ‪WN ( 0, σ 2‬‬

‫‪E ⎡⎣( zt − µ ) at ⎤⎦ = E ( at at ) = σ 2‬‬

‫‪ at‬إذا‬

‫‪k = 0 : γ 0 = E ⎡⎣( zt − µ ) at ⎤⎦ = σ 2‬‬
‫‪k = 1: γ 1 = E ⎡⎣( zt −1 − µ ) at ⎤⎦ = 0‬‬

‫وذﻟﻚ ﻷن‬

‫‪zt −1 − µ = at −1‬‬
‫‪E ⎡⎣( zt −1 − µ ) at ⎤⎦ = E ( at −1at ) = 0‬‬

‫ﻓﻲ اﻟﺤﻘﻴﻘﺔ ﻓﺈن‬

‫‪٢١‬‬

zt −k − µ = at −k , k = 1, 2,K

E ⎡⎣( zt −k − µ ) at ⎦⎤ = E ( at −k at ) = 0, k = 1, 2,K
‫أي‬

:١ ‫ﻗﺎﻋﺪة‬
⎧σ , k = 0
E ⎡⎣ ( zt −k − µ ) at ⎤⎦ = E ( at −k at ) = ⎨
⎩ 0, k = 1, 2,..
2

‫أي‬

γ0 =σ
γ k = 0, k = ±1, ±2,K
2

:‫وﺗﻮﺿﻊ ﻋﻠﻲ ﺷﻜﻞ داﻟﻲ‬

⎧σ , k = 0
γk = ⎨
2

⎩ 0, k ≠ 0

‫ ﻧﺠﺪ‬γ 0 = σ 2 ‫وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻲ‬

γ k ⎧1, k = 0
=⎨
γ 0 ⎩0, k ≠ 0
:‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬
Autocorrelation function of Constant Mean Model
1.0

Autocorr

ρk =

0.5

0.0
0

1

2

3

4

5

6

7

8

9

Lag

‫ ﻧﺠﺪ‬١١ ‫ﻧﺸﺘﻖ اﻵن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻣﻦ اﻟﺘﻌﺮﻳﻒ‬

٢٢

φ00 = 1, by definition
φ11 = ρ1 , by definition
φ11 = 0

1

ρ1 1 0
ρ2 0 0
=
=0
ρ1 1 0

ρ1

1

1

ρ1

ρ1
ρ2

1

1

φ22 =

φ33 =

ρ1

0 1

1

ρ1 1 0
ρ2 0 1
ρ3 0 0
=
ρ2 1 0
ρ1 0 1

ρ1

1

ρ1
ρ1

1

ρ1
ρ2

0
0
0
=0
0
0
0 0 1

M

1

ρ1

ρ1

1
M

M

ρ k −1 ρ k −2
ρ1
1
ρ1
1

φkk =

M

M

ρ k −1 ρ k −2

L
L
M
L
L
L
M
L

ρ1
ρ2

1
0
M
0
=
1
0
M
0

M

ρk
ρ k −1
ρ k −2
M
1

0
1
M
0
0
1
M
0

L
L
M
L
L
L
M
L

0
0
M
0 0
= = 0, k = 2,3,L
0 1
0
M
1
:‫وﺗﻮﺿﻊ ﻋﻠﻲ ﺷﻜﻞ داﻟﻲ‬

⎧1, k = 0
⎩0, k ≠ 0

φkk = ⎨

Partial Autocorrelation function of Constant Mean Model

:‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬

PACF

1.0

0.5

0.0
0

1

2

3

4

5

6

7

8

9

Lag
‫اﻟﺒﻴﻀﺎء اﻻ ﻓﻲ ان ﻟﻪ ﻣﺘﻮﺳﻂ ﻏﻴﺮ ﺻﻔﺮي‬
‫ ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﺜﺎﺑﺖ ﻻﻳﻔﺘﺮق ﻋﻦ ﻧﻤﻮذج اﻟﻀﺠﺔ‬:‫ﻣﻼﺣﻈﺔ‬

٢٣

‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻻوﻟﻲ )‪ARMA(١،٠) = AR(١‬‬
‫وهﻮ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬
‫‪φ1 ( B ) zt = δ + θ 0 ( B ) at‬‬

‫‪(1 − φ1B ) zt = δ + at‬‬

‫) ‪WN ( 0,σ 2‬‬

‫‪zt = δ + φ1 zt −1 + at , at‬‬

‫آﺎﻟﻨﻤﻮذج اﻟﺴﺎﺑﻖ ﺳﻮف ﻧﻮﺟﺪ اﻟﺘﻮﻗﻊ )اﻟﻤﺘﻮﺳﻂ( وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‪:‬‬

‫‪(1 − φ1B ) zt = δ + at‬‬

‫‪+ (1 − φ1 B ) at‬‬

‫‪δ‬‬

‫‪−1‬‬

‫) ‪(1 − φ1‬‬

‫‪δ‬‬

‫‪−1‬‬
‫⎤ ‪+ E ⎡(1 − φ1 B ) at‬‬
‫⎦‬
‫⎣ ) ‪(1 − φ1‬‬

‫= ‪zt‬‬

‫= ) ‪E ( zt‬‬

‫اﻟﺤﺪ اﻟﺜﺎﻧﻲ ﻓﻲ اﻟﻄﺮف اﻷﻳﻤﻦ هﻮ‬

‫⎤ ⎞ ‪⎡⎛ ∞ j j‬‬
‫‪−1‬‬
‫⎡‬
‫⎤‬
‫⎥ ‪E (1 − φ1 B ) at = E ⎢⎜ ∑φ1 B ⎟ at‬‬
‫⎣‬
‫⎦‬
‫‪⎢⎣⎝ j =0‬‬
‫⎦⎥ ⎠‬
‫ﻹدﺧﺎل اﻟﺘﻮﻗﻊ ﻋﻠﻲ اﻟﻤﺠﻤﻮع اﻟﻶﻧﻬﺎﺋﻲ ﻳﺠﺐ ان ﺗﻜﻮن اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻶﻧﻬﺎﺋﻴﺔ ∞ < ‪B j‬‬

‫‪j‬‬

‫∞‬

‫‪∑φ‬‬

‫‪1‬‬

‫ﻣﺘﻘﺎرﺑﺔ وذﻟﻚ‬

‫‪j =0‬‬

‫ﻳﺘﺤﻘﻖ إذا آﺎﻧﺖ ‪ φ1 < 1‬وذﻟﻚ إذا اﻋﺘﺒﺮﻧﺎ اﻟﻌﺎﻣﻞ ‪ B‬اﻵن ﻳﻠﻌﺐ دور ﻣﺘﻐﻴﺮ ﻣﺮآﺐ ‪Complex‬‬
‫‪ Variable‬ﻟﻪ اﻟﺸﻜﻞ ‪ B = a + ib‬وﻟﻪ اﻟﻘﻴﺎس ‪ B = 1‬ﻓﻲ اﻟﺤﻘﻴﻘﺔ ﻻﺑﺪ ان ﻧﺘﻄﻠﺐ ان ﺗﻜﻮن ﺟﺰور او‬
‫اﺻﻔﺎر ‪ (1 − φ1 B ) = 0‬ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة أي ‪ B > 1‬أي‬
‫‪1 − φ1 B = 0‬‬
‫‪1‬‬

‫‪φ1‬‬
‫‪> 1 ⇒ φ1 < 1‬‬

‫‪1‬‬

‫‪φ1‬‬

‫=‪B‬‬

‫⇒‪B >1‬‬

‫وهﺬا هﻮ ﺷﺮط اﻹﺳﺘﻘﺮار‪ .‬ﻧﻌﻮد اﻟﻲ اﻟﻌﻼﻗﺔ‬

‫⎤ ⎞ ‪⎡⎛ ∞ j‬‬
‫‪−1‬‬
‫⎥ ‪E ⎡(1 − φ1 B ) at ⎤ = E ⎢⎜ ∑φ1 B j ⎟ at‬‬
‫⎣‬
‫⎦‬
‫⎥⎦ ⎠‬
‫‪⎣⎢⎝ j =0‬‬
‫⎞ ‪⎡⎛ ∞ j j‬‬
‫⎤‬
‫⎥ ) ‪= ⎢⎜ ∑φ1 B ⎟ E ( at‬‬
‫‪⎢⎣⎝ j =0‬‬
‫⎦⎥‬
‫⎠‬
‫‪=0, ∀t‬‬
‫وﻳﻜﻮن‬

‫‪٢٤‬‬

‫‪δ‬‬

‫) ‪(1 − φ1‬‬
‫او‬

‫= ) ‪E ( zt‬‬

‫‪δ‬‬

‫=‪µ‬‬

‫) ‪(1 − φ1‬‬
‫) ‪∴δ = µ (1 − φ1‬‬

‫وﺑﺎﻟﺘﻌﻮﻳﺾ ﻋﻦ ‪ δ‬ﻓﻲ ﺻﻴﻐﺔ اﻟﻨﻤﻮذج ﻧﺠﺪ‬

‫‪zt = δ + φ1 zt −1 + at‬‬

‫‪= µ (1 − φ1 ) + φ1 zt −1 + at‬‬
‫‪= µ + φ1 ( zt −1 − µ ) + at‬‬
‫ﻧﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻓﻲ ‪ zt −k − µ‬وﻧﺄﺧﺬ اﻟﺘﻮﻗﻊ أي‬

‫‪( zt − µ ) − φ1 ( zt −1 − µ ) = at‬‬

‫‪E ⎡⎣( zt −k − µ )( zt − µ ) ⎤⎦ − φ1E ⎡⎣( zt −k − µ )( zt −1 − µ ) ⎤⎦ = E ⎡⎣( zt −k − µ ) at ⎤⎦ , k = 0, ±1, ±2,L‬‬
‫أي‬

‫‪γ k − φ1γ k −1 = E ⎡⎣( zt −k − µ ) at ⎤⎦ , k = 0, ±1, ±2,L‬‬

‫وذﻟﻚ ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٨‬و ﺗﺤﻞ هﺬﻩ اﻟﻌﻼﻗﺔ ﺗﻜﺮارﻳﺎ آﻤﺎ ﻳﻠﻲ‪:‬‬
‫⎦⎤ ‪k = 0 : γ 0 − φ1γ 1 = E ⎡⎣( zt − µ ) at‬‬
‫ﻹﻳﺠﺎد اﻟﻄﺮف اﻷﻳﻤﻦ ﻧﻘﻮم ﺑﺎﻟﺘﺎﻟﻲ‪:‬‬
‫) ‪E ⎡⎣ at ( zt − µ ) ⎤⎦ − φ1 E ⎡⎣ at ( zt −1 − µ ) ⎤⎦ = E ( at at‬‬
‫‪E ⎡⎣ at ( zt − µ ) ⎤⎦ − φ1 × ( 0 ) = σ 2‬‬
‫‪∴ E ⎡⎣ at ( zt − µ ) ⎤⎦ = σ 2‬‬
‫إذا‬

‫‪γ 0 − φ1γ 1 = σ‬‬
‫‪k = 1: γ 1 − φ1γ 0 = E ⎡⎣( zt −1 − µ ) at ⎤⎦ = 0‬‬
‫‪2‬‬

‫ﻓﻲ اﻟﺤﻘﻴﻘﺔ‬

‫ﺑﻘﺴﻤﺔ اﻟﻤﻌﺎدﻟﺔ اﻷﺧﻴﺮة ﻋﻠﻲ ‪γ 0‬‬

‫‪γ k − φ1γ k −1 = 0, k = 1, 2,L‬‬
‫ﻧﺠﺪ‬

‫‪ρ k − φ1 ρ k −1 = 0, k = 1, 2,L‬‬

‫أو‬

‫وﺑﻤﺎ ان ‪ρ 0 = 1‬‬

‫‪ρ k = φ1 ρ k −1 , k = 1, 2,L‬‬
‫ﻓﺈن‪:‬‬

‫‪ρ1 = φ1 ρ 0 = φ1‬‬
‫‪ρ 2 = φ1 ρ1 = φ12‬‬

‫‪٢٥‬‬

‫‪M‬‬

‫‪ρ k = φ1k‬‬
‫أو ﺑﺸﻜﻞ داﻟﺔ‬

‫‪ρ k = φ , k = 0, ±1, ±2,L‬‬
‫وذﻟﻚ ﻷن ‪ ρ − k = ρ k , ∀k‬ﺳﻮف ﻧﻨﻈﺮ ﻣﻦ اﻵن وﺻﺎﻋﺪا ﻟﻠﺸﻖ اﻟﻤﻮﺟﺐ ﻣﻦ ‪ ρ k‬أي‬
‫‪k‬‬
‫‪ρ k = φ1 , k = 0,1, 2,L‬‬
‫هﺬﻩ اﻟﺪاﻟﺔ ﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪ -١‬ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪φ1 > 0‬‬
‫‪k‬‬
‫‪1‬‬

‫‪Autocorrelation function of AR(1) Model‬‬
‫‪0.5‬‬
‫‪0.4‬‬

‫‪ACF‬‬

‫‪0.3‬‬
‫‪0.2‬‬
‫‪0.1‬‬
‫‪0.0‬‬
‫‪10‬‬

‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪4‬‬

‫‪5‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫‪Lag‬‬

‫‪ -٢‬ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪φ1 < 0‬‬
‫‪Autocorrelation function of AR(1) Model‬‬
‫‪0.3‬‬
‫‪0.2‬‬
‫‪0.1‬‬
‫‪0.0‬‬

‫‪-0.2‬‬
‫‪-0.3‬‬
‫‪-0.4‬‬
‫‪-0.5‬‬
‫‪10‬‬

‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪4‬‬

‫‪5‬‬

‫‪Lag‬‬

‫ﻧﺸﺘﻖ اﻵن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬
‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ١١‬ﻧﺠﺪ‬

‫‪٢٦‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫‪ACF‬‬

‫‪-0.1‬‬

φ00 = 1, by definition
φ11 = ρ1 = φ1 , by definition
1

φ22 =

ρ1
1

ρ1

1 φ1
ρ1
ρ 2 φ1 φ12
0
=
=
=0
1 φ1 1 − φ12
ρ1
1
φ1 1

M
1

ρ1

ρ1

1
M

M

ρ
ρ k −2
φkk = k −1
1
ρ1
1
ρ1
M

M

ρ k −1 ρ k −2

L
L
L
L
L
L
L
L

ρ1
ρ2
M

1

φ1

φ1

1
M

M

ρk
φ
=
ρ k −1
1
ρ k −2
φ1

φ1k −2
φ1 L

k −1
1

M
1

φ

L φ1
L φ12
L M
L φ1k

M

φ

k −1
1

φ1k
L φ1k −1

1
M

L
L

k −2
1

=

0
>0

M
1

‫ وﻧﻜﺘﺐ‬φ1 ‫ﻣﺤﺪدة اﻟﺒﺴﻂ ﺗﺴﺎوي ﺻﻔﺮا ﻷن اﻟﻌﺎﻣﻮد اﻷﺧﻴﺮ ﻳﺴﺎوي اﻟﻌﺎﻣﻮد اﻷول ﻣﻀﺮوﺑﺎ ﻓﻲ‬
:‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻋﻠﻲ اﻟﺸﻜﻞ اﻟﺪاﻟﻲ‬
⎧ 1, k = 0

φkk = ⎨φ1 , k = 1
⎪ 0, k ≥ 2

:‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬
φ1 > 0 ‫ ﻋﻨﺪﻣﺎ ﺗﻜﻮن‬-١
Partial Autocorrelation function of AR(1) Model
0.5

PACF

0.4
0.3
0.2
0.1
0.0
0

1

2

3

4

5

6

7

8

9

10

Lag

φ1 < 0 ‫ ﻋﻨﺪﻣﺎ ﺗﻜﻮن‬-٢

٢٧

‫‪Partial Autocorrelation function of AR(1) Model‬‬
‫‪0.0‬‬
‫‪-0.1‬‬

‫‪-0.3‬‬

‫‪PACF‬‬

‫‪-0.2‬‬

‫‪-0.4‬‬
‫‪-0.5‬‬

‫‪10‬‬

‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪4‬‬

‫‪5‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫‪Lag‬‬

‫ﻣﻼﺣﻈﺔ‪ :‬داﺋﻤﺎ ﻻﺗﺮﺳﻢ أي ﻣﻦ ‪ ρ 0 = 1‬او ‪ φ00 = 1‬ﻓﻲ اﻷﺷﻜﺎل اﻟﺒﻴﺎﻧﻴﺔ‪.‬‬

‫ﻣﻨﺎﻗﺸﺔ اﻟﻨﻤﻮذج‪:‬‬
‫‪-١‬‬
‫‪-٢‬‬
‫‪-٣‬‬
‫‪-٤‬‬

‫ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪) φ1 < 1‬ﺷﺮط اﻹﺳﺘﻘﺮار( ﻓﺈن ) ‪ E ( zt ) = δ (1 − φ1‬وهﻮﺛﺎﺑﺖ ﻟﺠﻤﻴﻊ ﻗﻴﻢ ‪t‬‬
‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ داﻟﺔ ﻟﻠﺘﺨﻠﻒ ‪ k‬ﻓﻘﻂ وﻻﺗﻌﺘﻤﺪ ﻋﻠﻲ اﻟﺰﻣﻦ ‪t‬‬
‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﺗﺘﺨﺎﻣﺪ اﺳﻴﺎ ﻓﻲ إﺗﺠﺎﻩ واﺣﺪ إﺑﺘﺪاءا ﻣﻦ ‪ ρ1‬ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪ φ1 > 0‬وﺗﺘﺨﺎﻣﺪ‬
‫اﺳﻴﺎ ﻣﺘﺮددة ﺑﻴﻦ اﻟﻘﻴﻢ اﻟﻤﻮﺟﺒﺔ واﻟﺴﺎﻟﺒﺔ ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪φ1 < 0‬‬
‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻬﺎ ﻗﻴﻤﺔ واﺣﺪة ﻏﻴﺮ ﺻﻔﺮﻳﺔ ) ﻣﻊ ﻋﺪم اﻟﻨﻈﺮ اﻟﻲ ‪ ( φ00‬وﻳﻜﻮن‬
‫إﺗﺠﺎهﻬﺎ ﺣﺴﺐ إﺷﺎرة ‪ φ1‬وﻣﻘﺪارهﺎ ﻳﺴﺎوي ‪φ1‬‬

‫ﺛﺎﻟﺜﺎ‪ :‬ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ )‪: ARMA(٢،٠) = AR(٢‬‬
‫وﻳﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬
‫‪φ 2 ( B ) zt = δ + θ 0 ( B ) a t‬‬
‫‪= δ + at‬‬
‫) ‪WN ( 0,σ 2‬‬

‫‪(1 − φ B + φ B ) z‬‬
‫‪2‬‬

‫‪t‬‬

‫‪2‬‬

‫‪1‬‬

‫‪zt = δ + φ1 zt −1 + φ2 zt −2 + at , at‬‬

‫آﺎﻟﺴﺎﺑﻖ ﻧﻮﺟﺪ اﻟﻤﺘﻮﺳﻂ وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‪:‬‬
‫‪(1 − φ1B − φ2 B ) zt = δ + at‬‬
‫‪2‬‬

‫‪−1‬‬
‫‪δ‬‬
‫‪+ (1 − φ1 B − φ2 B 2 ) at‬‬
‫) ‪(1 − φ1 − φ2‬‬
‫‪−1‬‬
‫‪δ‬‬
‫⎤ ‪+ E ⎡(1 − φ1 B − φ2 B 2 ) at‬‬
‫⎦⎥‬
‫⎣⎢ ) ‪(1 − φ1 − φ2‬‬

‫‪٢٨‬‬

‫= ‪zt‬‬

‫= ) ‪E ( zt‬‬

‫∞ ⎛‬
‫⎞‬
‫اﻟﺤﺪ اﻟﺜﺎﻧﻲ ﻓﻲ اﻟﻄﺮف اﻷﻳﻤﻦ ﻣﺠﻤﻮع ﻻﻧﻬﺎﺋﻲ ﻋﻠﻰ اﻟﺸﻜﻞ ⎟ ‪ E ⎜ ∑ψ j at − j‬وﻟﻜﻲ ﻧﺪﺧﻞ اﻟﺘﻮﻗﻊ‬
‫‪⎝ j =0‬‬
‫⎠‬
‫∞‬

‫داﺧﻞ اﻟﺘﺠﻤﻴﻊ اﻟﻼﻧﻬﺎﺋﻲ ﻻﺑﺪ ان ﺗﻜﻮن‬

‫‪∑ψ a‬‬

‫‪j t− j‬‬

‫ﻣﺘﻘﺎرﺑﺔ ﻓﻲ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺮﺑﻊ وهﺬا ﻳﺘﺤﻘﻖ إذا‬

‫‪j =0‬‬

‫وﻓﻘﻂ إذا آﺎن ∞ <‬

‫‪2‬‬
‫‪j‬‬

‫∞‬

‫‪∑ψ‬‬

‫وهﺬا ﻳﺘﺤﻘﻖ إذا ﺣﻘﻘﺖ ﻣﻌﺎﻟﻢ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﺸﺮوط اﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪j =0‬‬

‫‪φ2 − φ1 < 1‬‬
‫‪φ2 + φ1 < 1‬‬
‫‪−1 < φ2 < 1‬‬
‫واﻟﺘﻲ ﺗﺴﻤﻲ ﺑﺸﺮوط اﻹﺳﺘﻘﺮار ) هﺬﻩ اﻟﺸﺮوط ﺗﻨﺘﺞ اﻳﻀﺎ ﻣﻦ آﻮن ﺟﺰور او أﺻﻔﺎر‬
‫‪ (1 − φ1 B − φ2 B 2 ) = 0‬ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة ( ‪ .‬إذا ﺗﺤﻘﻘﺖ ﺷﺮوط اﻹﺳﺘﻘﺮار ﻓﺈن‬
‫‪−1‬‬
‫‪−1‬‬
‫‪E ⎡(1 − φ1 B − φ2 B 2 ) at ⎤ = ⎡(1 − φ1 B − φ2 B 2 ) E ( at ) ⎤ = 0, ∀t‬‬
‫⎢⎣ ⎦⎥‬
‫⎦⎥‬
‫⎢⎣‬

‫وﻳﻜﻮن‬

‫‪δ‬‬
‫) ‪(1 − φ1 − φ2‬‬

‫و ﺑﺎﻟﺘﻌﻮﻳﺾ ﻋﻦ ‪ δ‬ﻓﻲ ﺻﻴﻐﺔ اﻟﻨﻤﻮذج ﻧﺠﺪ‬

‫= ) ‪µ = E ( zt‬‬

‫‪δ = (1 − φ1 − φ2 ) µ‬‬
‫‪zt = (1 − φ1 − φ2 ) µ + φ1 zt −1 + φ2 zt −2 + at‬‬
‫‪= µ + φ1 ( zt −1 − µ ) + φ2 ( zt −2 − µ ) + at‬‬

‫‪( zt − µ ) − φ1 ( zt −1 − µ ) − φ2 ( zt −2 − µ ) = at‬‬
‫ﻧﻀﺮب اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻓﻲ ‪ zt −k − µ‬وﻧﺄﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‪:‬‬

‫⎦⎤ ) ‪E ⎡⎣( zt − µ )( zt −k − µ ) − φ1 ( zt −1 − µ )( zt −k − µ ) − φ2 ( zt −2 − µ )( zt −k − µ‬‬
‫‪= E ⎡⎣ at ( zt −k − µ ) ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫أي‬
‫⎦⎤ ) ‪E ⎡⎣( zt − µ )( zt −k − µ ) ⎦⎤ − φ1 E ⎡⎣ ( zt −1 − µ )( zt −k − µ ) ⎤⎦ − φ2 E ⎡⎣( zt −2 − µ )( zt −k − µ‬‬
‫‪= E ⎡⎣ at ( zt −k − µ ) ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫أو‬

‫‪γ k − φ1γ k −1 − φ2γ k −2 = E ⎡⎣ at ( zt −k − µ ) ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫وذﻟﻚ ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٨‬اﻵن ﻧﺤﻞ هﺬﻩ اﻟﻌﻼﻗﺔ ﺗﻜﺮارﻳﺎ آﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪٢٩‬‬

‫‪k = 0 : γ 0 − φ1γ −1 − φ2γ −2 = E ⎡⎣ at ( zt − µ ) ⎤⎦ = σ 2 ⇒ γ 0 = φ1γ 1 − φ2γ 2 + σ 2‬‬
‫وذﻟﻚ ﻣﻦ ﻗﺎﻋﺪة ‪١‬‬
‫‪k = 1: γ 1 − φ1γ 0 − φ2γ 1 = 0 ⇒ γ 1 = φ1γ 0 − φ2γ 1‬‬
‫‪k = 2 : γ 2 − φ1γ 1 − φ2γ 0 = 0 ⇒ γ 2 = φ1γ 1 − φ2γ 0‬‬
‫وﺑﺸﻜﻞ ﻋﺎم‬
‫‪k ≥ 1: γ k = φ1γ k −1 + φ2γ k −2‬‬
‫ﺑﻘﺴﻤﺔ اﻟﻄﺮﻓﻴﻦ ﻋﻠﻲ ‪ γ 0‬ﻧﺠﺪ‬
‫‪ρ k = φ1 ρ k −1 + φ2 ρ k −2 , k = 1, 2, ...‬‬
‫) ﻣﻼﺣﻈﺔ‪ :‬ﺑﻮﺿﻊ اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻋﻠﻲ اﻟﺸﻜﻞ ‪ρ k − φ1 ρ k −1 − φ2 ρ k −2 = 0, k = 1, 2,...‬‬
‫ﻧﺠﺪ اﻧﻬﺎ ﻣﻌﺎدﻟﺔ ﻓﺮوﻗﻴﺔ ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ واﻟﺘﻲ ﻳﻤﻜﻦ ﺣﻠﻬﺎ ﺑﺸﻜﻞ ﻣﻐﻠﻖ ﺑﺈﺳﺘﺨﺪام ﻃﺮق ﺣﻞ‬
‫اﻟﻤﻌﺎدﻻت اﻟﻔﺮوﻗﻴﺔ وﻟﻜﻦ هﺬا ﺧﺎرج ﻧﻄﺎق اﻟﻤﻘﺮر اﻟﺤﺎﻟﻲ(‬
‫ﺳﻮف ﻧﺤﻞ اﻟﻌﻼﻗﺔ اﻟﺴﺎﺑﻘﺔ ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺘﻜﺮارﻳﺔ واﻟﺘﻲ ﺗﺤﺘﺎج اﻟﻲ ﻗﻴﻤﺘﻴﻦ اوﻟﻴﺘﻴﻦ‪:‬‬
‫‪1 − ρ0 = 1‬‬
‫‪φ‬‬
‫‪2 − ρ1 = φ1 ρ 0 + φ2 ρ −1 ⇒ ρ1 = 1‬‬
‫‪1 − φ2‬‬
‫وﻣﻨﻬﺎ ﻧﺠﺪ‬
‫‪2‬‬
‫‪φ‬‬
‫‪ρ 2 = φ1 ρ1 + φ2 ρ0 ⇒ ρ 2 = 1 + φ2‬‬
‫‪1 − φ2‬‬
‫وهﻜﺬا اﻟﺦ…‬
‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ هﻲ ﻟﺪوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻌﻤﻠﻴﺔ ) ‪AR ( 2‬‬
‫‪ -١‬اﻟﺸﻜﻞ )‪φ1 = 0.4, φ2 = 0.4 (١‬‬
‫‪ -٢‬اﻟﺸﻜﻞ )‪φ1 = 1.5, φ2 = −0.8 (٢‬‬
‫‪ -٣‬اﻟﺸﻜﻞ )‪φ1 = 0.5, φ2 = −0.6 (٣‬‬
‫ﺷﻜﻞ )‪(١‬‬
‫‪ACF‬‬
‫‪0.7‬‬
‫‪0.6‬‬
‫‪0.5‬‬

‫‪0.3‬‬
‫‪0.2‬‬
‫‪0.1‬‬
‫‪0.0‬‬
‫‪20‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٣٠‬‬

‫‪0‬‬

‫‪ACF‬‬

‫‪0.4‬‬

(٢) ‫ﺷﻜﻞ‬
ACF
1.0

ACF

0.5

0.0

-0.5

0

10

20

Lag

(٣) ‫ﺷﻜﻞ‬
ACF

ACF

0.5

0.0

-0.5

0

10

20

Lag

:‫ آﺎﻟﺘﺎﻟﻲ‬AR ( 2 ) ‫اﻵن ﻧﺸﺘﻖ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻌﻤﻠﻴﺔ‬

φ00 = 1, by definition
φ11 = ρ1 , by definition
1 ρ1
ρ1 ρ 2 ρ 2 − ρ12
=
≠0
φ22 =
1 ρ1
1 − ρ12
ρ1 1

٣١

‫‪=0‬‬

‫‪ρ1 = φ1 + φ2 ρ1‬‬
‫‪ρ 2 = φ1 ρ1 + φ2‬‬
‫‪ρ3 = φ1 ρ 2 + φ2 ρ1‬‬

‫‪ρ1‬‬
‫‪1‬‬

‫‪ρ1‬‬

‫‪>0‬‬

‫‪ρ1‬‬
‫‪1‬‬
‫‪ρ2‬‬
‫‪ρ1‬‬
‫‪ρ3 ρ 2‬‬
‫=‬
‫‪ρ2‬‬
‫‪ρ1‬‬

‫‪ρ1‬‬

‫‪1‬‬

‫‪1‬‬

‫‪ρ1‬‬
‫‪ρ2‬‬

‫‪ρ1‬‬
‫‪ρ1‬‬
‫‪1‬‬

‫‪ρ1‬‬

‫‪1‬‬

‫= ‪φ33‬‬

‫‪ρ1‬‬
‫‪ρ2‬‬

‫‪1‬‬
‫وذﻟﻚ ﻷن اﻟﻌﻤﻮد اﻷﺧﻴﺮ ﻓﻲ ﻣﺤﺪدة اﻟﺒﺴﻂ هﻮ ﺗﺮآﻴﺐ ﺧﻄﻲ ﻣﻦ اﻟﻌﻤﻮدﻳﻦ اﻷول واﻟﺜﺎﻧﻲ‪ ،‬آﺬﻟﻚ‬
‫‪1‬‬
‫‪ρ1 L ρ1‬‬
‫‪1‬‬
‫‪ρ1 L‬‬
‫‪ρ1 = φ1 ρ 0 + φ2 ρ1‬‬
‫‪ρ1‬‬
‫‪1 L ρ2‬‬
‫‪ρ1‬‬
‫‪1 L‬‬
‫‪ρ 2 = φ1 ρ1 + φ2 ρ 0‬‬
‫‪M‬‬
‫‪M‬‬
‫‪L M‬‬
‫‪M‬‬
‫‪M‬‬
‫‪L‬‬
‫‪M‬‬
‫‪ρ‬‬
‫‪ρ k −2 L ρ k‬‬
‫‪ρ‬‬
‫‪ρ k −2 L ρ k = φ1 ρ k −1 + φ2 ρ k −2‬‬
‫‪φkk = k −1‬‬
‫‪= k −1‬‬
‫‪= 0, k = 3, 4,...‬‬
‫‪1‬‬
‫‪ρ1 L ρ k −1‬‬
‫‪>0‬‬
‫‪1 L ρ k −2‬‬
‫‪ρ1‬‬
‫‪M‬‬
‫‪M‬‬
‫‪L‬‬
‫‪M‬‬
‫‪ρ k −1 ρ k −2 L 1‬‬
‫وذﻟﻚ اﻳﻀﺎ ﻟﻨﻔﺲ اﻟﺴﺒﺐ اﻟﺴﺎﺑﻖ‪ .‬إذا‬

‫‪k =0‬‬
‫‪⎧ 1,‬‬
‫‪⎪ ρ,‬‬
‫‪k =1‬‬
‫‪1‬‬
‫⎪⎪‬
‫‪φkk = ⎨ ρ 2 − ρ12‬‬
‫‪⎪ 1− ρ2 , k = 2‬‬
‫‪1‬‬
‫⎪‬
‫‪k ≥3‬‬
‫‪⎪⎩ 0,‬‬

‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ هﻲ ﻟﺪوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻌﻤﻠﻴﺔ ) ‪AR ( 2‬‬
‫‪ -٤‬اﻟﺸﻜﻞ )‪φ1 = 0.4, φ2 = 0.4 (٤‬‬
‫‪ -٥‬اﻟﺸﻜﻞ )‪φ1 = 1.5, φ2 = −0.8 (٥‬‬
‫‪ -٦‬اﻟﺸﻜﻞ )‪φ1 = 0.5, φ2 = −0.6 (٦‬‬
‫ﺷﻜﻞ )‪(٤‬‬

‫‪PACF‬‬
‫‪0.7‬‬
‫‪0.6‬‬
‫‪0.5‬‬

‫‪0.3‬‬
‫‪0.2‬‬
‫‪0.1‬‬
‫‪0.0‬‬
‫‪20‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٣٢‬‬

‫‪0‬‬

‫‪PACF‬‬

‫‪0.4‬‬

‫ﺷﻜﻞ )‪(٥‬‬
‫‪PACF‬‬
‫‪1‬‬

‫‪PACF‬‬

‫‪0‬‬

‫‪-1‬‬

‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺷﻜﻞ )‪(٦‬‬
‫‪PACF‬‬
‫‪0.3‬‬
‫‪0.2‬‬
‫‪0.1‬‬

‫‪-0.2‬‬

‫‪PACF‬‬

‫‪0.0‬‬
‫‪-0.1‬‬

‫‪-0.3‬‬
‫‪-0.4‬‬
‫‪-0.5‬‬
‫‪-0.6‬‬
‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫راﺑﻌﺎ‪ :‬ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻷوﻟﻲ )‪: ARMA(٠،١) = MA(١‬‬
‫وﺗﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬
‫‪φ0 ( B ) zt = δ + θ1 ( B ) at‬‬

‫) ‪WN ( 0,σ 2‬‬

‫‪zt = δ + (1 − θ1 B ) at‬‬
‫‪zt = δ + at − θ1at −1 , at‬‬

‫اﻵن ﻧﻮﺟﺪ اﻟﻤﺘﻮﺳﻂ وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‪:‬‬
‫‪E ( zt ) = E (δ + at − θ1at −1 ) = δ‬‬
‫‪∴µ = δ‬‬

‫وﻧﻜﺘﺐ اﻟﻨﻤﻮذج‬
‫ﺑﻀﺮب هﺬﻩ اﻟﻤﻌﺎدﻟﺔ ﻓﻲ ‪ zt −k − µ‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬

‫‪zt − µ = at − θ1at −1‬‬

‫‪E ⎡⎣( zt − µ )( zt −k − µ ) ⎤⎦ = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1E ⎡⎣( zt −k − µ ) at −1 ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫‪٣٣‬‬

‫او‬

‫‪γ k = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1 E ⎡⎣( zt −k − µ ) at −1 ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫وﺑﺤﻠﻬﺎ ﺗﻜﺮارﻳﺎ‬

‫⎦⎤ ‪k = 0 : γ 0 = E ⎡⎣( zt − µ ) at ⎤⎦ − θ1 E ⎡⎣( zt − µ ) at −1‬‬

‫ﻧﻮﺟﺪ آﻞ ﻣﻦ ⎦⎤ ‪ E ⎡⎣( zt − µ ) at‬و ⎦⎤ ‪ E ⎡⎣( zt − µ ) at −1‬آﺎﻵﺗﻲ‪:‬‬
‫‪E ⎡⎣( zt − µ ) at ⎤⎦ = E ( at at ) − θ1 E ( at −1at ) = σ 2‬‬
‫‪E ⎡⎣( zt − µ ) at −1 ⎤⎦ = E ( at at −1 ) − θ1E ( at −1at −1 ) = −θ1σ 2‬‬

‫) ‪∴γ 0 = σ 2 − θ1 ( −θ1σ 2 ) = σ 2 (1 + θ12‬‬

‫⎦⎤ ‪k = 1: γ 1 = E ⎡⎣( zt −1 − µ ) at ⎤⎦ − θ1 E ⎡⎣( zt −1 − µ ) at −1‬‬
‫‪−θ1‬‬
‫‪γ1‬‬
‫=‬
‫‪γ 0 1 + θ12‬‬
‫وذﻟﻚ ﺑﺈﺳﺘﺨﺪام اﻟﻘﺎﻋﺪة ‪١‬‬

‫= ‪∴ γ 1 = −θ1σ 2 ⇒ ρ1‬‬

‫⎦⎤ ‪k = 2 : γ 2 = E ⎡⎣( zt −2 − µ ) at ⎤⎦ − θ1 E ⎡⎣( zt −2 − µ ) at −1‬‬
‫‪∴ γ 2 = 0 ⇒ ρ2 = 0‬‬

‫أﻳﻀﺎ ﻣﻦ ﻗﺎﻋﺪة ‪ ١‬وﺑﺸﻜﻞ ﻋﺎم ﻓﺈن‬

‫‪k ≥ 2 : γ k = 0 ⇒ ρk = 0‬‬
‫وهﻜﺬا ﻓﺈن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻨﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻷوﻟﻲ )‪ MA (1‬هﻲ ﻋﻠﻲ‬
‫اﻟﺸﻜﻞ‪:‬‬
‫‪⎧ 1,‬‬
‫‪k =0‬‬
‫⎪‬
‫‪⎪ −θ‬‬
‫‪ρk = ⎨ 1 2 , k = 1‬‬
‫‪⎪ 1 + θ1‬‬
‫‪⎪⎩ 0‬‬
‫‪k ≥2‬‬
‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪ -١‬ﻋﻨﺪﻣﺎ ‪θ1 = 0.8‬‬
‫‪ACF‬‬
‫‪0.0‬‬
‫‪-0.1‬‬

‫‪-0.3‬‬
‫‪-0.4‬‬
‫‪-0.5‬‬
‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٣٤‬‬

‫‪ACF‬‬

‫‪-0.2‬‬

‫‪ -٢‬ﻋﻨﺪﻣﺎ ‪θ1 = −0.8‬‬
‫‪ACF‬‬
‫‪0.5‬‬
‫‪0.4‬‬

‫‪0.2‬‬

‫‪ACF‬‬

‫‪0.3‬‬

‫‪0.1‬‬
‫‪0.0‬‬

‫‪20‬‬

‫‪10‬‬

‫‪0‬‬

‫‪Lag‬‬

‫اﻵن ﻧﺸﺘﻖ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻷوﻟﻲ )‪MA (1‬‬
‫‪φ00 = 1, by definition‬‬

‫‪φ11 = ρ1 , by definition‬‬
‫‪1 ρ1‬‬
‫‪1 ρ1‬‬
‫) ‪−θ12 (1 − θ12‬‬
‫‪ρ1 ρ 2‬‬
‫‪ρ1 0‬‬
‫‪− ρ12‬‬
‫‪−θ12‬‬
‫= ‪φ22‬‬
‫=‬
‫=‬
‫=‬
‫=‬
‫‪1 ρ1‬‬
‫‪1 − ρ12‬‬
‫‪1 − ρ12 1 + θ12 + θ14‬‬
‫‪1 − θ16‬‬
‫‪ρ1 1‬‬
‫‪ρ1‬‬
‫‪1 ρ1 ρ1‬‬
‫‪ρ2‬‬
‫‪ρ1 1 0‬‬
‫) ‪−θ13 (1 − θ12‬‬
‫‪ρ3‬‬
‫‪0 ρ1 0‬‬
‫‪ρ13‬‬
‫=‬
‫=‬
‫=‬
‫‪ρ2‬‬
‫‪1 ρ1 0 1 − 2 ρ12‬‬
‫‪1 − θ18‬‬
‫‪ρ1‬‬
‫‪ρ1 1 ρ1‬‬
‫‪1‬‬
‫‪0 ρ1 1‬‬
‫وﺑﺸﻜﻞ ﻋﺎم‬
‫‪k >0‬‬

‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪ -١‬ﻋﻨﺪﻣﺎ ﻋﻨﺪﻣﺎ ‪θ1 = −0.8‬‬

‫‪٣٥‬‬

‫‪ρ1‬‬

‫‪1‬‬

‫‪1‬‬

‫‪ρ1‬‬
‫‪ρ2‬‬

‫‪ρ1‬‬
‫‪ρ1‬‬
‫‪1‬‬

‫‪ρ1‬‬

‫‪(1 − θ ) ,‬‬
‫‪2‬‬
‫‪1‬‬

‫)‪2 k +1‬‬

‫‪1‬‬

‫= ‪φ33‬‬

‫‪ρ1‬‬
‫‪ρ2‬‬
‫‪k‬‬
‫‪1‬‬

‫‪−θ‬‬

‫( ‪1 − θ1‬‬

‫= ‪φkk‬‬

‫‪PACF‬‬
‫‪0.5‬‬
‫‪0.4‬‬
‫‪0.3‬‬

‫‪0.1‬‬
‫‪0.0‬‬

‫‪PACF‬‬

‫‪0.2‬‬

‫‪-0.1‬‬
‫‪-0.2‬‬
‫‪-0.3‬‬
‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪ -٢‬ﻋﻨﺪﻣﺎ ‪θ1 = 0.8‬‬
‫‪PACF‬‬
‫‪0.0‬‬
‫‪-0.1‬‬

‫‪-0.3‬‬

‫‪PACF‬‬

‫‪-0.2‬‬

‫‪-0.4‬‬
‫‪-0.5‬‬
‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺧﺎﻣﺴﺎ‪ :‬ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ )‪: ARMA(٠،٢) = MA(٢‬‬
‫وﺗﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬
‫‪φ 0 ( B ) zt = δ + θ 2 ( B ) a t‬‬
‫‪zt = δ + (1 − θ1 B − θ 2 B 2 ) at‬‬

‫) ‪WN ( 0,σ 2‬‬

‫‪zt = δ + at − θ1at −1 − θ 2 at −2 , at‬‬

‫اﻵن ﻧﻮﺟﺪ اﻟﻤﺘﻮﺳﻂ وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‪:‬‬
‫‪E ( zt ) = E (δ + at − θ1at −1 − θ 2 at −2 ) = δ‬‬
‫‪∴µ = δ‬‬

‫‪٣٦‬‬

‫وﻧﻜﺘﺐ اﻟﻨﻤﻮذج‬
‫ﺑﻀﺮب هﺬﻩ اﻟﻤﻌﺎدﻟﺔ ﻓﻲ ‪ zt −k − µ‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬

‫‪zt − µ = at − θ1at −1 − θ 2 at −2‬‬

‫⎦⎤ ‪E ⎡⎣( zt − µ )( zt −k − µ ) ⎤⎦ = E ⎡⎣ ( zt −k − µ ) at ⎤⎦ − θ1 E ⎡⎣ ( zt −k − µ ) at −1‬‬
‫‪− θ 2 E ⎡⎣( zt −k − µ ) at −2 ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫او‬
‫‪− µ ) at −2 ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫‪γ k = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1E ⎡⎣( zt −k − µ ) at −1 ⎤⎦ − θ 2 E ⎡⎣( zt −k‬‬

‫وﺑﺤﻠﻬﺎ ﺗﻜﺮارﻳﺎ ﻧﺠﺪ‬
‫‪2‬‬

‫‪γ 0 = (1 + θ + θ ) σ‬‬
‫‪2‬‬
‫‪2‬‬

‫‪2‬‬
‫‪1‬‬

‫‪γ 1 = ( −θ1 + θ1θ 2 ) σ 2‬‬
‫‪γ 2 = −θ 2σ 2‬‬
‫‪γ k = 0, k > 2‬‬

‫وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻲ ‪ γ 0‬ﻧﺠﺪ‬

‫‪−θ1 + θ1θ 2‬‬
‫‪1 + θ12 + θ 22‬‬

‫= ‪ρ1‬‬

‫‪−θ 2‬‬
‫‪1 + θ12 + θ 22‬‬

‫= ‪ρ2‬‬

‫‪ρ k = 0, k > 2‬‬
‫وﺗﻜﺘﺐ ﻋﻠﻲ ﺷﻜﻞ داﻟﺔ‬
‫‪1,‬‬
‫‪k =0‬‬
‫⎧‬
‫‪⎪ −θ + θ θ‬‬
‫‪⎪ 1 2 1 22 , k = 1‬‬
‫‪⎪ 1 + θ1 + θ 2‬‬
‫⎨ = ‪ρk‬‬
‫‪⎪ −θ 2‬‬
‫‪, k =2‬‬
‫‪⎪1 + θ12 + θ 22‬‬
‫⎪‬
‫‪0,‬‬
‫‪k>2‬‬
‫⎩‬

‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ هﻲ ﻟﺪوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻌﻤﻠﻴﺔ ) ‪MA ( 2‬‬
‫‪ -٧‬اﻟﺸﻜﻞ )‪θ1 = 0.4, θ 2 = 0.4 (٧‬‬
‫‪ -٨‬اﻟﺸﻜﻞ )‪θ1 = 1.5, θ 2 = −0.8 (٨‬‬
‫‪ -٩‬اﻟﺸﻜﻞ )‪θ1 = 0.5, θ 2 = −0.6 (٩‬‬

‫‪٣٧‬‬

(٧) ‫ﺷﻜﻞ‬

ACF
0.0

ACF

-0.1

-0.2

-0.3

0

10

20

Lag

(٨) ‫ﺷﻜﻞ‬

ACF
0.2
0.1
0.0

-0.2
-0.3
-0.4
-0.5
-0.6
-0.7
0

10

20

Lag

(٩) ‫ﺷﻜﻞ‬
ACF
0.4
0.3

0.2
0.1

ACF

ACF

-0.1

0.0
-0.1
-0.2
-0.3
-0.4
-0.5
0

10

Lag

٣٨

20

‫ﻣﻦ اﻟﺼﻌﺐ ﺟﺪا إﻳﺠﺎد ﺷﻜﻞ ﻣﻐﻠﻖ ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ‬
‫اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ ) ‪ MA ( 2‬وﻟﻬﺬا ﺳﻮف ﻧﺴﺘﺨﺪم ﺗﻌﺮﻳﻒ ‪ ١١‬ب ﻟﺤﺴﺎﺑﻬﺎ ورﺳﻤﻬﺎ ﺗﻜﺮارﻳﺎ ﻟﻘﻴﻢ‬
‫اﻟﻤﻌﺎﻟﻢ اﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ -١٠‬اﻟﺸﻜﻞ )‪θ1 = 0.4, θ 2 = 0.4 (١٠‬‬
‫‪ -١١‬اﻟﺸﻜﻞ )‪θ1 = 1.5, θ 2 = −0.8 (١١‬‬
‫‪ -١٢‬اﻟﺸﻜﻞ )‪θ1 = 0.5, θ 2 = −0.6 (١٢‬‬

‫ﺷﻜﻞ )‪(١٠‬‬

‫‪PACF‬‬
‫‪0.0‬‬

‫‪-0.1‬‬

‫‪PACF‬‬

‫‪-0.2‬‬

‫‪-0.3‬‬

‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺷﻜﻞ )‪(١١‬‬
‫‪PACF‬‬
‫‪0.2‬‬
‫‪0.1‬‬
‫‪0.0‬‬
‫‪-0.1‬‬

‫‪-0.3‬‬
‫‪-0.4‬‬
‫‪-0.5‬‬
‫‪-0.6‬‬
‫‪-0.7‬‬

‫‪20‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٣٩‬‬

‫‪0‬‬

‫‪PACF‬‬

‫‪-0.2‬‬

‫ﺷﻜﻞ )‪(١٢‬‬
‫‪PACF‬‬
‫‪0.4‬‬
‫‪0.3‬‬
‫‪0.2‬‬
‫‪0.1‬‬

‫‪PACF‬‬

‫‪0.0‬‬
‫‪-0.1‬‬
‫‪-0.2‬‬
‫‪-0.3‬‬
‫‪-0.4‬‬
‫‪-0.5‬‬
‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺳﺎدﺳﺎ‪ :‬ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪-‬اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ )‪: ARMA(١،١‬‬
‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‪:‬‬

‫‪φ1 ( B ) zt = δ + θ1 ( B ) at‬‬

‫‪(1 − φ1B ) zt = δ + (1 − θ1B ) at‬‬
‫‪WN ( 0, σ 2 ) , φ1 ≠ θ1‬‬

‫‪zt − φ1 zt −1 = δ + at − θ1at −1‬‬
‫‪zt = δ + φ1 zt −1 + at − θ1at −1 , at‬‬

‫ﺷﺮط اﻹﺳﺘﻘﺮار ‪ φ1 < 1‬وﺷﺮط اﻹﻧﻘﻼب ‪ θ1 < 1‬وهﻨﺎك ﺷﺮط ﺁﺧﺮ ﻳﺴﻤﻰ ﺷﺮط اﻹﻣﺘﺴﺎخ‬
‫‪ Degeneracy Condition‬وهﻮ ‪ φ1 ≠ θ1‬وهﺬا اﻟﺸﺮط ﻳﻀﻤﻦ ﻋﺪم إﻣﺘﺴﺎخ اﻟﻨﻤﻮذج إﻟﻰ ﻧﻤﻮذج‬
‫أﻗﻞ درﺟﺔ ﻓﻔﻲ ﺣﺎﻟﺔ آﻮن ‪ φ1 = θ1‬ﻓﻤﻦ اﻟﻌﻼﻗﺔ ‪ (1 − φ1B ) zt = δ + (1 − θ1B ) at‬وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻰ‬
‫) ‪ (1 − φ1B‬ﻧﺠﺪ أن اﻟﻨﻤﻮذج ﻳﺼﺒﺢ ‪ zt = δ ′ + at‬ﺣﻴﺚ‬
‫ﻧﻮﺟﺪ اﻟﻤﺘﻮﺳﻂ آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫‪δ‬‬
‫‪1 − φ1‬‬

‫= ‪ δ ′‬وهﻮ ) ‪ARMA ( 0,0‬‬

‫‪(1 − φ1B ) zt = δ + (1 − θ1B ) at‬‬
‫‪(1 − θ1B ) a‬‬
‫‪δ‬‬
‫= ‪zt‬‬
‫‪+‬‬
‫‪t‬‬
‫) ‪1 − φ1 (1 − φ1B‬‬
‫‪(1 − θ1B ) E a‬‬
‫‪δ‬‬
‫= ) ‪E ( zt‬‬
‫‪+‬‬
‫)‪( t‬‬
‫) ‪1 − φ1 (1 − φ1B‬‬

‫وذﻟﻚ ﻷن ‪ φ1 < 1‬وهﻜﺬا‬
‫‪δ‬‬

‫أي‬

‫‪δ‬‬
‫‪1 − φ1‬‬

‫‪1 − φ1‬‬

‫= ) ‪E ( zt‬‬

‫= ‪ E ( zt ) = µ‬أو ) ‪ δ = µ (1 − φ1‬وﺑﺎﻟﺘﻌﻮﻳﺾ ﻋﻦ ‪ δ‬ﻧﺠﺪ‬
‫‪zt = µ (1 − φ1 ) + φ1 zt −1 + at − θ1at −1‬‬

‫‪( zt − µ ) − φ1 ( zt −1 − µ ) = at − θ1at −1‬‬
‫وﺑﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ ﺑﺎﻟﺤﺪ ‪ ( zt −k − µ ) , k = 0, ±1, ±2,...‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻟﻠﻄﺮﻓﻴﻦ ﻧﺠﺪ‬
‫‪E ⎡⎣( zt −k − µ )( zt − µ )⎤⎦ − φ1E ⎡⎣ ( zt −k − µ )( zt −1 − µ )⎤⎦ = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1E ⎡⎣( zt −k − µ ) at −1 ⎤⎦ ,‬‬
‫‪k = 0, ±1, ±2,...‬‬

‫وﻣﻨﻬﺎ‬
‫‪٤٠‬‬

γ k − φ1γ k −1 = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1E ⎡⎣ ( zt −k − µ ) at −1 ⎤⎦ , k = 0, ±1, ±2,...

‫وﺑﺤﻠﻬﺎ ﺗﻜﺮارﻳﺎ ﻧﺠﺪ‬

k = 0 γ 0 − φ1γ 1 = E ⎡⎣ ( zt − µ ) at ⎤⎦ − θ1E ⎡⎣ ( zt − µ ) at−1 ⎤⎦

‫ ﺑﻀﺮب اﻟﻌﻼﻗﺔ‬E ⎡⎣( zt − µ ) at −1 ⎤⎦ ‫ و‬E ⎡⎣( zt − µ ) at ⎤⎦ ‫ﻧﻮﺟﺪ اﻵن آﻞ ﻣﻦ‬

( zt − µ ) − φ1 ( zt −1 − µ ) = at − θ1at −1

‫ وأﺧﺬ اﻟﺘﻮﻗﻊ‬at −1 ‫ و‬at ‫ﻓﻲ آﻞ ﻣﻦ‬
E ⎡⎣ ( zt − µ ) at ⎤⎦ − φ1E ⎡⎣ ( zt −1 − µ ) at ⎤⎦ = E [at at ] − θ1 E [at −1at ]

‫ ﻧﺠﺪ‬١ ‫وﻣﻦ اﻟﻘﺎﻋﺪة‬

E ⎡⎣( zt − µ ) at ⎤⎦ − φ1 ( 0 ) = σ − θ1 ( 0 )
2

E ⎡⎣( zt − µ ) at ⎤⎦ = σ 2
E ⎡⎣( zt − µ ) at −1 ⎤⎦ − φ1E ⎡⎣( zt −1 − µ ) at −1 ⎤⎦ = E [at at −1 ] − θ1E [at −1at −1 ]

‫و‬

E ⎡⎣( zt − µ ) at −1 ⎤⎦ − φ1σ 2 = 0 − θ1σ 2
∴ E ⎡⎣( zt − µ ) at −1 ⎤⎦ = σ 2 (φ1 − θ1 )

‫وﺑﺎﻟﺘﻌﻮﻳﺾ ﻓﻲ اﻟﺼﻴﻐﺔ اﻟﺴﺎﺑﻘﺔ ﻧﺠﺪ‬
k = 0 γ 0 − φ1γ 1 = σ 2 − θ1σ 2 (φ1 − θ1 )
∴γ 0 − φ1γ 1 = σ 2 ⎡⎣1 − θ1 (φ1 − θ1 )⎤⎦
k = 1 γ 1 − φ1γ 0 = E ⎡⎣ ( zt −1 − µ ) at ⎤⎦ − θ1 E ⎡⎣ ( zt −1 − µ ) at −1 ⎤⎦

‫و‬

∴ γ 1 − φ1γ 0 = −θ1σ 2
k = 2 γ 2 − φ1γ 1 = E ⎡⎣ ( zt −2 − µ ) at ⎤⎦ − θ1 E ⎡⎣ ( zt −2 − µ ) at −1 ⎤⎦ = 0

‫و‬

∴ k ≥ 2 γ k − φ1γ k −1 = 0

‫وﻣﻦ اﻟﻤﻌﺎدﻻت‬

γ 0 − φ1γ 1 = σ ⎡⎣1 − θ1 (φ1 − θ1 )⎤⎦
2

‫و‬
γ 1 − φ1γ 0 = −θ1σ

2

‫ﻧﺠﺪ‬
γ0 =
γ1 =

ρ1 =

1 + θ − 2φ1θ1 2
σ
1 − φ12
2
1

(1 − φ1θ1 )(φ1 − θ1 ) σ 2
1 − φ12

‫وﻣﻦ اﻟﻌﻼﻗﺘﻴﻦ اﻟﺴﺎﺑﻘﺘﻴﻦ ﻧﺠﺪ‬

γ 1 (1 − φ1θ1 )(φ1 − θ1 )
=
1 + θ12 − 2φ1θ1
γ0

٤١

‫وﻣﻦ اﻟﻌﻼﻗﺔ‬

‫‪γ k − φ1γ k −1 = 0, k ≥ 2‬‬

‫وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻰ ‪ γ 0‬ﻧﺠﺪ‬
‫‪ρ k − φ1 ρ k −1 = 0, k ≥ 2‬‬
‫وﻳﻤﻜﻦ ﺣﻞ هﺬﻩ اﻟﻤﻌﺎدﻟﺔ ﺗﻜﺮارﻳﺎ ﻟﺠﻤﻴﻊ ﻗﻴﻢ ‪ k ≥ 2‬ﺑﺈﺳﺘﺨﺪام اﻟﻘﻴﻢ اﻷوﻟﻴﺔ ‪ ρ0 = 1‬و‬
‫) ‪(1 − φ1θ1 )(φ1 − θ1‬‬
‫= ‪ ρ1‬ﻓﻤﺜﻼ‬
‫‪1 + θ12 − 2φ1θ1‬‬
‫‪ρ2 = φ1 ρ1‬‬
‫) ‪(1 − φ1θ1 )(φ1 − θ1‬‬
‫‪ρ 2 = φ1‬‬
‫‪1 + θ12 − 2φ1θ1‬‬
‫‪ρ3 = φ1 ρ 2‬‬
‫) ‪(1 − φ1θ1 )(φ1 − θ1‬‬
‫‪ρ3 = φ12‬‬
‫‪1 + θ12 − 2φ1θ1‬‬

‫وهﻜﺬا‪.‬‬
‫ﻧﻜﺘﺐ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻨﻤﻮذج )‪ ARMA (1,1‬ﻋﻠﻰ اﻟﺸﻜﻞ‬
‫‪1,‬‬
‫‪k =0‬‬
‫⎧‬
‫⎪‬
‫) ‪⎪ (1 − φ1θ1 )(φ1 − θ1‬‬
‫‪, k =1‬‬
‫⎨ = ‪ρk‬‬
‫‪2‬‬
‫‪⎪ 1 + θ1 − 2φ1θ1‬‬
‫⎩⎪‬
‫‪k≥2‬‬
‫‪φ1 ρ k −1‬‬

‫ﺷﻜﻞ ‪ ١٣‬ﻳﻌﻄﻲ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻘﻴﻢ ‪φ1 = 0.9,θ1 = −0.5‬‬

‫ﺷﻜﻞ)‪(١٣‬‬

‫) ‪A C F o f A R M A (1 ,1‬‬
‫‪1 .0‬‬
‫‪0 .9‬‬
‫‪0 .8‬‬
‫‪0 .7‬‬
‫‪0 .6‬‬
‫‪0 .4‬‬
‫‪0 .3‬‬
‫‪0 .2‬‬
‫‪0 .1‬‬
‫‪0 .0‬‬

‫‪15‬‬

‫‪5‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺷﻜﻞ ‪ ١٤‬ﻳﻌﻄﻲ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻘﻴﻢ ‪φ1 = −0.9,θ1 = −0.5‬‬

‫‪٤٢‬‬

‫‪0‬‬

‫‪C1‬‬

‫‪0 .5‬‬

‫ﺷﻜﻞ)‪(١٤‬‬
‫) ‪A C F o f A R M A (1 ,1‬‬
‫‪0 .5‬‬

‫‪C1‬‬

‫‪0 .0‬‬

‫‪-0 .5‬‬

‫‪10‬‬

‫‪15‬‬

‫‪0‬‬

‫‪5‬‬

‫‪Lag‬‬

‫ﻧﻼﺣﻆ ان داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻨﻤﻮذج )‪ ARMA (1,1‬ﺗﺘﺨﺎﻣﺪ اﺳﻴﺎ ﻓﻲ إﺗﺠﺎﻩ واﺣﺪ أو ﻣﺘﺮدد ﺑﻴﻦ‬
‫اﻟﻘﻴﻢ اﻟﻤﻮﺟﺒﺔ واﻟﺴﺎﻟﺒﺔ وهﻲ ﻓﻲ هﺬا ﺗﺸﺒﻪ ﺗﻤﺎﻣﺎ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻨﻤﻮذج )‪ AR (1‬ﻣﺎﻋﺪى ان‬
‫اﻟﺘﺨﺎﻣﺪ ﻳﺒﺪأ ﻣﻦ ‪ ) ρ1‬ﺑﺮهﻦ أن ‪( ρ k = φ1k −1 ρ1 , k ≥ 2‬‬
‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ‪ φkk‬ﺗﺤﺴﺐ ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ١١‬أو ‪١١‬ب آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪١١‬ب ﻧﻮﺟﺪ ‪ φkk‬ﺗﻜﺮارﻳﺎ‬
‫‪φ00 = 1, by definition‬‬

‫) ‪(1 − φ1θ1 )(φ1 − θ1‬‬
‫‪1 + θ12 − 2φ1θ1‬‬

‫= ‪φ11 = ρ1‬‬

‫‪ρ 2 − φ11 ρ1‬‬
‫‪1 − φ11 ρ1‬‬
‫‪ρ −φ ρ −φ ρ‬‬
‫‪φ33 = 3 21 2 22 1 , φ21 = φ11 − φ22φ11‬‬
‫‪1 − φ21 ρ1 − φ22 ρ 2‬‬

‫= ‪φ22‬‬

‫وهﻜﺬا ﺗﺤﺴﺐ ﺑﻘﻴﺔ اﻟﻘﻴﻢ ﺗﻜﺮارﻳﺎ‪.‬‬
‫ﻓﻤﺜﻼ ﻟﻠﻘﻴﻢ ‪ φ1 = 0.9,θ1 = −0.5‬ﻧﺠﺪ‬

‫‪φ11 = 0.944186 φ22 = -0.384471 φ33 = 0.183710‬‬
‫‪φ44 = -0.908462 φ55 = 0.452979 φ66 = -0.226337‬‬
‫‪φ77 = 0.113154 φ88 = -0.565702 φ99 = 0.282834‬‬

‫وﻧﺮﺳﻢ هﺬﻩ اﻟﻘﻴﻢ ﻓﻲ ﺷﻜﻞ ‪١٥‬‬
‫ﺷﻜﻞ ‪١٥‬‬
‫) ‪P A C F o f A R M A (1 ,1‬‬
‫‪1 .0‬‬

‫‪0 .5‬‬

‫‪C2‬‬
‫‪0 .0‬‬

‫‪15‬‬

‫‪5‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٤٣‬‬

‫‪0‬‬

‫ﺷﻜﻞ ‪ ١٦‬ﻳﺒﻴﻦ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻘﻴﻢ ‪φ1 = −0.9,θ1 = −0.5‬‬

‫ﺷﻜﻞ ‪١٦‬‬
‫) ‪P A C F o f A R M A (1 ,1‬‬
‫‪0 .3‬‬
‫‪0 .2‬‬
‫‪0 .1‬‬
‫‪0 .0‬‬

‫‪-0 .2‬‬

‫‪C2‬‬

‫‪-0 .1‬‬

‫‪-0 .3‬‬
‫‪-0 .4‬‬
‫‪-0 .5‬‬
‫‪-0 .6‬‬

‫‪15‬‬

‫‪5‬‬

‫‪10‬‬

‫‪0‬‬

‫‪Lag‬‬

‫ﻧﻼﺣﻆ ان داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج )‪ ARMA (1,1‬ﺗﺘﺨﺎﻣﺪ اﺳﻴﺎ ﻓﻲ إﺗﺠﺎﻩ واﺣﺪ أو ﻣﺘﺮدد‬
‫ﺑﻴﻦ اﻟﻘﻴﻢ اﻟﻤﻮﺟﺒﺔ واﻟﺴﺎﻟﺒﺔ وهﻲ ﻓﻲ هﺬا ﺗﺸﺒﻪ ﺗﻤﺎﻣﺎ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج )‪MA (1‬‬
‫ﻣﺎﻋﺪى ان اﻟﺘﺨﺎﻣﺪ ﻳﺒﺪأ ﺑﻌﺪ اﻟﻘﻴﻤﺔ اﻷوﻟﻴﺔ ‪. φ11 = ρ1‬‬

‫ﺧﻮاص ﻧﻤﺎذج )‪: ARMA(p,q‬‬
‫أوﻻ‪ :‬ﻧﻤﻮذج )‪AR(p‬‬
‫وﻳﺘﻤﻴﺰ ﺑﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪ -١‬داﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﺗﻤﺘﺪ ﻻﻧﻬﺎﺋﻴﺎ وﺗﺘﻜﻮن ﻣﻦ ﺧﻠﻴﻂ ﻣﻦ اﻟﺘﺨﺎﻣﺪات اﻻﺳﻴﺔ واﻟﺘﺨﺎﻣﺪات اﻟﺠﻴﺒﻴﺔ‪.‬‬
‫‪ -٢‬داﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﺟﺰﺋﻲ ﺗﺘﻜﻮن ﻣﻦ أﺻﻔﺎر ﻟﻘﻴﻢ اﻟﺘﺨﻠﻔﺎت ‪ k > p‬أي‬
‫‪φ11 = φ22 = φ33 = L = φ pp ≠ 0‬‬
‫‪φ p +1, p +1 = φ p +2, p + 2 = L = 0‬‬

‫وﻳﺴﻤﻰ هﺬا ﻗﻄﻌﺎ ﻓﻲ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺑﻌﺪ اﻟﺘﺨﻠﻒ ‪. k > p‬‬
‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﻤﻮذج )‪: MA(q‬‬
‫وﻳﺘﻤﻴﺰ ﺑﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪ -١‬داﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﺗﺘﻜﻮن ﻣﻦ أﺻﻔﺎر ﻟﻘﻴﻢ اﻟﺘﺨﻠﻔﺎت ‪ k > q‬أي‬
‫‪ρ1 = ρ 2 = ρ3 = L = ρ q ≠ 0‬‬
‫‪ρ q+1,q+1 = ρ q+2,q +2 = L = 0‬‬

‫وﻳﺴﻤﻰ هﺬا ﻗﻄﻌﺎ ﻓﻲ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﺑﻌﺪ اﻟﺘﺨﻠﻒ ‪. k > q‬‬
‫‪ -٢‬داﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﺟﺰﺋﻲ ﺗﻤﺘﺪ ﻻﻧﻬﺎﺋﻴﺎ وﺗﺘﻜﻮن ﻣﻦ ﺧﻠﻴﻂ ﻣﻦ اﻟﺘﺨﺎﻣﺪات اﻻﺳﻴﺔ واﻟﺘﺨﺎﻣﺪات‬
‫اﻟﺠﻴﺒﻴﺔ‪.‬‬
‫ﻻﺣﻆ اﻹزدواﺟﻴﺔ ‪ Duality‬ﺑﻴﻦ ﻧﻤﻮذﺟﻲ ‪ AR‬و ‪.MA‬‬
‫ﺛﺎﻟﺜﺎ‪ :‬اﻟﻨﻤﻮذج اﻟﻤﺨﺘﻠﻂ )‪:ARMA(p,q‬‬
‫وﻳﺘﻤﻴﺰ ﺑﺎﻟﺘﺎﻟﻲ‪:‬‬
‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻨﻤﻮذج اﻟﻤﺨﺘﻠﻂ ﺗﻤﺘﺪ ﻻﻧﻬﺎﺋﻴﺎ وﺗﺘﻜﻮن ﻣﻦ ﺧﻠﻴﻂ ﻣﻦ‬
‫اﻟﺘﺨﺎﻣﺪات اﻻﺳﻴﺔ واﻟﺘﺨﺎﻣﺪات اﻟﺠﻴﺒﻴﺔ اﻟﺘﻲ ﺗﻨﺘﻬﻲ إﻟﻰ اﻟﺼﻔﺮ آﻠﻤﺎ زاد اﻟﺘﺨﻠﻒ ‪ . k‬ﻋﻨﺪﻣﺎ ﺗﻜﻮن‬
‫‪ k > q − p‬ﻓﺈن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﺗﺘﺤﺪد ﻣﻦ ﺟﺰء اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻟﻠﻨﻤﻮذج و ﻋﻨﺪﻣﺎ ﺗﻜﻮن‬
‫‪ k > p − q‬ﻓﺈن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﺘﺤﺪد ﻣﻦ ﺟﺰء اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻟﻠﻨﻤﻮذج‪.‬‬
‫‪٤٤‬‬

‫اﻟﻔﺼﻞ اﻟﺜﺎﻟﺚ‬
‫ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ‪Nonstationar Time‬‬
‫‪:Series Models‬‬
‫اوﻻ‪ :‬ﻋﺪم اﻹﺳﺘﻘﺮار ﻓﻲ اﻟﻤﺘﻮﺳﻂ‪:‬‬
‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٦‬ﻹﺳﺘﻘﺮار ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻧﺮى ان اﻟﺸﺮط اﻷول ﻟﻺﺳﺘﻘﺮار‬
‫‪ E ( zt ) = µ = constant ∀t‬ﻳﺘﻄﻠﺐ أن ﻳﻜﻮن ﻣﺘﻮﺳﻂ اﻟﻤﺘﺴﻠﺴﻠﺔ ﺛﺎﺑﺖ ﻋﻠﻰ ﻃﻮل اﻟﺰﻣﻦ‪ ،‬ﻓﻤﺜﻼ‬
‫ﻟﻨﻤﻮذج اﻹﻧﺠﺮاف اﻟﺨﻄﻲ‬
‫) ∞ ‪WN ( 0, σ 2 ) , b0 , b1 ∈ ( −∞,‬‬

‫‪zt = b0 + b1t + at , at‬‬

‫ﻧﺠﺪ ان اﻟﻤﺘﻮﺳﻂ هﻮ‬

‫‪E ( z )t = b0 + b1t‬‬

‫وهﻮ ﻏﻴﺮ ﺛﺎﺑﺖ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﺰﻣﻦ‪ ،‬اي ان ﺷﺮط اﻹﺳﺘﻘﺮار اﻷول ﻏﻴﺮ ﻣﺘﺤﻘﻖ ﻓﻲ هﺬﻩ اﻟﺤﺎﻟﺔ‪.‬‬
‫ﻟﻨﺤﺎول اﻟﺘﺤﻮﻳﻞ ‪ ∇zt‬وذﻟﻚ ﺑﺘﻄﺒﻴﻖ ﻋﺎﻣﻞ اﻟﺘﻔﺮﻳﻖ ﻋﻠﻰ اﻟﻨﻤﻮذج ﻧﺠﺪ‬

‫‪wt = ∇zt = zt − zt −1 = b0 + b1t + at − b0 − b1 ( t − 1) − at −1‬‬
‫‪=b1 + at − at −1 = b1 + ct‬‬

‫) ‪WN ( 0,ν 2‬‬

‫) ﺗﻤﺮﻳﻦ‪ :‬أوﺟﺪ اﻟﻌﻼﻗﺔ ﺑﻴﻦ ‪ ν 2‬و ‪( σ 2‬‬
‫اﻵن ﻧﺠﺪ ﻣﺘﻮﺳﻂ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺠﺪﻳﺪة ‪wt‬‬

‫‪∴ wt = b1 + ct , ct‬‬

‫‪E ( wt ) = b1 = constant ∀t‬‬

‫أي ان ﺗﻄﺒﻴﻖ اﻟﺘﺤﻮﻳﻞ ) ‪ ∇ = (1 − B‬ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ) أي أﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻷول‬
‫ﻟﻠﻤﺘﺴﻠﺴﻠﺔ( ﺣﻮﻟﻬﺎ إﻟﻰ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة‪.‬‬
‫آﻤﺜﺎل ﺁﺧﺮ ﻧﻤﻮذج اﻹﻧﺠﺮاف اﻟﺘﺮﺑﻴﻌﻲ‬
‫) ∞ ‪WN ( 0, σ 2 ) , b0 , b1 , b2 ∈ ( −∞,‬‬

‫‪zt = b0 + b1t + b2t 2 + at , at‬‬

‫ﺑﺈﻳﺠﺎد اﻟﻤﺘﻮﺳﻂ‬
‫‪2‬‬

‫‪E ( zt ) = b0 + b1t + b2t‬‬

‫وهﻮ ﻳﻌﺘﻤﺪ ﻋﻠﻰ اﻟﺰﻣﻦ‪ ،‬أي ان اﻟﻨﻤﻮذج ﻏﻴﺮ ﻣﺴﺘﻘﺮ‪ .‬ﺑﺄﺧﺬ اﻟﺘﺤﻮﻳﻞ ‪ ) ∇ 2 zt‬أﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻟﺜﺎﻧﻲ(‬
‫ﻧﺠﺪ‬

‫) ‪∇2 zt = ∇ 2 ( b0 + b1t + b2t 2 + at‬‬

‫) ‪+ b1t + b2t 2 + at‬‬

‫‪(1 − 2 B + B ) z = (1 − 2 B + B )( b‬‬
‫‪2‬‬

‫‪2‬‬

‫‪0‬‬

‫‪t‬‬

‫‪wt = {b0 − 2b0 + b0 } + {b1t − 2b1 ( t − 1) + b1 ( t − 2 )} +‬‬

‫}‬

‫‪− 2b2 ( t − 1) + b2 ( t − 2 ) +‬‬
‫‪2‬‬

‫‪2‬‬

‫‪2‬‬

‫‪{b t‬‬
‫‪2‬‬

‫} ‪{at − 2at −1 + at −2‬‬
‫} ‪= 2b2 + {at − 2at −1 + at −2‬‬
‫) ‪WN ( 0,τ 2‬‬

‫وهﻜﺬا‬

‫‪٤٥‬‬

‫‪=b′ + ht , ht‬‬

‫) ‪WN ( 0,τ 2‬‬

‫‪wt = ∇2 zt = b′ + ht , ht‬‬

‫‪E ( wt ) = b′ = constant ∀t‬‬

‫أي ان ﺗﻄﺒﻴﻖ اﻟﺘﺤﻮﻳﻞ ‪) ∇2‬أي اﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻟﺜﺎﻧﻲ( ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ﺣﻮﻟﻬﺎ اﻟﻰ‬
‫ﻣﺴﺘﻘﺮة‪.‬‬
‫‪2‬‬
‫‪2‬‬
‫) ﺗﻤﺮﻳﻦ‪ :‬أوﺟﺪ اﻟﻌﻼﻗﺔ ﺑﻴﻦ ‪ τ‬و ‪.( σ‬‬
‫ﺑﺸﻜﻞ ﻋﺎم إذا آﺎن اﻟﻨﻤﻮذج ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ∞ ‪WN ( 0, σ 2 ) , b0 , b1 ,L , bd ∈ ( −∞,‬‬

‫‪zt = b0 + b1t + L + bd t + at , at‬‬
‫‪d‬‬

‫ﻓﺈن اﻟﺘﺤﻮﻳﻞ ‪ ∇d zt‬ﻳﺤﻮﻟﻪ إﻟﻰ ﻧﻤﻮذج ﻣﺴﺘﻘﺮ‪ ،‬أي ان ‪ wt = ∇ d zt‬هﻮ ﻧﻤﻮذج ﻣﺴﺘﻘﺮ‪.‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١٦‬ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ‬
‫) ∞ ‪zt = b0 + b1t + L + bd t + at , at WN ( 0, σ 2 ) , b0 , b1 ,L , bd ∈ ( −∞,‬‬
‫اﻟﺘﺤﻮﻳﻞ ‪ ∇d zt‬وهﻮ اﻟﺘﻔﺮﻳﻖ ﻟﻠﺪرﺟﺔ ‪ d‬ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﻳﺤﻮﻟﻬﺎ إﻟﻰ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة‪.‬‬
‫‪d‬‬

‫ﺛﺎﻧﻴ ًﺎ‪ :‬ﻋﺪم اﻹﺳﺘﻘﺮار ﻓﻲ اﻟﺘﺒﺎﻳﻦ‪:‬‬
‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٦‬ﻹﺳﺘﻘﺮار ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ‪ ،‬اﻟﺸﺮط اﻟﺜﺎﻧﻲ‬

‫ﻳﺘﻄﻠﺐ أن ﻳﻜﻮن اﻟﺘﺒﺎﻳﻦ ﺛﺎﺑﺖ ﻟﺠﻤﻴﻊ ﻗﻴﻢ ‪. t‬‬
‫ﻓﻤﺜﻼ ﻟﻨﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ‬

‫‪V ( zt ) = γ 0 = constant ∀t‬‬

‫) ‪WN ( 0, σ 2‬‬

‫‪zt = zt −1 + at , at‬‬

‫ﻧﺠﺪ ﻣﻦ اﻟﺘﻌﻮﻳﺾ اﻟﻤﺘﻜﺮر‬
‫‪zt = a1 + a2 + L + at‬‬

‫وﺑﺈﺧﺬ اﻟﺘﻮﻗﻊ واﻟﺘﺒﺎﻳﻦ‬

‫‪E ( zt ) = 0 = constant ∀t‬‬
‫‪V ( zt ) = tσ 2‬‬

‫وﻧﻼﺣﻆ أن اﻟﺘﺒﺎﻳﻦ ﻳﻌﺘﻤﺪ ﻋﻠﻰ اﻟﺰﻣﻦ ‪. t‬‬
‫ﺑﺄﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻷول‬
‫‪wt = ∇zt = zt − zt −1 = at‬‬

‫وﺑﺈﺧﺬ اﻟﺘﻮﻗﻊ واﻟﺘﺒﺎﻳﻦ‬
‫‪E ( wt ) = 0 = constant ∀t‬‬
‫‪V ( wt ) = σ 2 = constant ∀t‬‬

‫إذًا اﻟﺘﻔﺮﻳﻖ اﻷول ﺣﻮل اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ﻓﻲ اﻟﺘﺒﺎﻳﻦ إﻟﻰ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة‪.‬‬
‫ﺑﺸﻜﻞ ﻋﺎم ﻟﻮ آﺎن اﻟﺘﺒﺎﻳﻦ داﻟﺔ ﻟﻤﺴﺘﻮى )ﻣﺘﻮﺳﻂ( ﻣﺘﻐﻴﺮ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪V ( zt ) = cf ( µt‬‬

‫ﺣﻴﺚ ‪ c > 0‬ﺛﺎﺑﺖ و )⋅( ‪ f‬داﻟﺔ ﻣﻌﺮوﻓﺔ ﺗﻌﻄﻰ ﻗﻴﻤﺔ ﻏﻴﺮ ﺳﺎﻟﺒﺔ و ‪ µt‬ﻣﺴﺘﻮى أو ﻣﺘﻮﺳﻂ ﻳﺘﻐﻴﺮ‬
‫ﻣﻊ اﻟﺰﻣﻦ و ﺑﺎﻟﺘﺎﻟﻲ ﻓﺈن اﻟﺘﺒﺎﻳﻦ ﻳﻌﺘﻤﺪ ﻋﻠﻰ اﻟﺰﻣﻦ وهﻨﺎ ﻧﺤﺎول إﻳﺠﺎد ﺗﺤﻮﻳﻞ ) ‪ T ( zt‬أي إﻳﺠﺎد داﻟﺔ‬
‫)⋅ ( ‪ T‬ﻹﺳﺘﻘﺮار اﻟﺘﺒﺎﻳﻦ‪.‬‬
‫‪٤٦‬‬

‫اﻟﺘﺤﻮﻳﻞ‬
‫‪λ‬‬

‫‪zt − 1‬‬

‫‪λ‬‬

‫= ) ‪yt = T ( zt‬‬

‫ﻳﻌﻄﻲ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﺘﺒﺎﻳﻦ ﺣﻴﺚ ) ∞ ‪ λ ∈ ( −∞,‬هﻮ ﻣﻌﻠﻢ اﻟﺘﺤﻮﻳﻞ‪ .‬اﻟﺠﺪول اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ‬
‫اﻟﻘﻴﻢ اﻷآﺜﺮ إﺳﺘﺨﺪاﻣﺎ ﻟﻠﻤﻌﻠﻢ ‪ λ‬ﻣﻊ اﻟﺘﺤﻮﻳﻼت اﻟﻤﻘﺎﺑﻠﺔ ﻟﻬﺎ‪:‬‬
‫‪1.0‬‬
‫‪zt‬‬

‫‪0.5‬‬

‫‪0.0‬‬

‫‪-0.5‬‬

‫‪zt‬‬

‫‪ln zt‬‬

‫‪1‬‬
‫‪zt‬‬

‫‪-0.1‬‬

‫‪λ‬‬

‫‪1‬‬
‫‪zt‬‬

‫‪yt‬‬

‫ﻣﺜﺎل‪:‬‬
‫اﻟﺸﻜﻞ)ا( ﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ واﻟﺘﺒﺎﻳﻦ ‪zt‬‬
‫‪O r ig in a l S e r ie s‬‬
‫‪400‬‬

‫‪300‬‬

‫)‪z(t‬‬
‫‪200‬‬

‫‪100‬‬
‫‪90‬‬

‫‪80‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪In d e x‬‬

‫اﻟﺸﻜﻞ)ب( اﻟﻤﺘﺴﻠﺴﻠﺔ ﺑﻌﺪ ﺗﺜﺒﻴﺖ اﻟﺘﺒﺎﻳﻦ ﺑﺈﺟﺮاء اﻟﺘﺤﻮﻳﻞ ‪yt = ln zt‬‬
‫‪T r a n s f o r m e d S e r ie s‬‬
‫‪6 .0‬‬

‫‪5 .5‬‬

‫)‪ln z(t‬‬
‫‪5 .0‬‬

‫‪90‬‬

‫‪80‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫اﻟﺸﻜﻞ)ج( اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺤﻮﻟﺔ ‪ yt‬ﺑﻌﺪ إﺟﺮاء اﻟﺘﻔﺮﻳﻖ اﻷول ‪∇yt = yt − yt −1‬‬

‫‪٤٧‬‬

‫‪10‬‬

‫‪In d e x‬‬

‫‪D if f e re n c e d a n d T ra n s f o rm e d S e rie s‬‬
‫‪0 .2‬‬

‫‪0 .1‬‬
‫)‪y(t)-y(t-1‬‬

‫‪0 .0‬‬

‫‪-0 .1‬‬

‫‪-0 .2‬‬

‫‪90‬‬

‫‪80‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪In d e x‬‬

‫ﻻﺣﻆ آﻴﻒ اﺻﺒﺤﺖ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة ﻓﻲ آﻞ ﻣﻦ اﻟﻤﺘﻮﺳﻂ واﻟﺘﺒﺎﻳﻦ‪.‬‬

‫ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪(p,d,q‬‬
‫‪Autoregressive-Integrated-Moving Average Models‬‬
‫)‪ARIMA(p,d,q‬‬
‫ﻳﻤﻜﻦ ﻧﻤﺬﺟﺔ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺴﺘﻘﺮة ‪ wt = ∇ d zt‬ﻋﻠﻰ ﺷﻜﻞ ﻧﻤﻮذج أﻧﺤﺪار ذاﺗﻲ‪-‬ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ‬
‫اﻟﺪرﺟﺔ ) ‪ ( p, q‬آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫) ‪WN ( 0, σ 2‬‬

‫أو‬

‫)‬

‫‪2‬‬

‫‪φ p ( B ) wt = φ p ( B ) ∇ d zt = δ + θ q ( B ) at , at‬‬

‫‪φ p ( B )(1 − B ) zt = δ + θ q ( B ) at , at WN ( 0,σ‬‬
‫‪d‬‬

‫وهﺬا اﻟﻨﻤﻮذج ﻳﺴﻤﻰ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ) ‪( p, d , q‬‬
‫ﺣﻴﺚ ) ∞ ‪ δ ∈ ( −∞,‬ﻣﻌﻠﻢ اﻹﻧﺠﺮاف‪.‬‬
‫أﻣﺜﻠﺔ ﻋﻠﻰ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ) ‪: ( p, d , q‬‬
‫اوﻻ‪ :‬ﻧﻤﻮذج اﻹﻧﺤﺪاراﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ ﻣﻦ اﻟﺪرﺟﺔ )‪ (١،١‬أو )‪ARIMA(١،١،٠‬‬
‫)‪: =ARI(١،١‬‬
‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪φ1 ( B )(1 − B ) zt = δ + θ 0 ( B ) at , at WN ( 0, σ 2‬‬

‫‪(1 − φ1B )(1 − B ) zt = δ + at‬‬
‫‪{1 − (φ1 + 1) B + φ1B 2 } zt = δ + at‬‬
‫أي‬
‫‪φ1 < 1‬‬

‫‪),‬‬

‫‪2‬‬

‫‪WN ( 0, σ‬‬

‫‪zt = δ + (φ1 + 1) zt −1 − φ1 zt −2 + at , at‬‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﻤﻮذج اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪(١،١‬‬
‫أو )‪: ARIMA(٠،١،١) = IMA(١،١‬‬
‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪φ0 ( B )(1 − B ) zt = δ + θ1 ( B ) at , at WN ( 0, σ 2‬‬

‫‪WN ( 0,σ 2 ) , θ1 < 1‬‬

‫‪at‬‬

‫‪(1 − B ) zt = δ + (1 − θ1B ) at ,‬‬

‫‪WN ( 0, σ 2 ) , θ1 < 1‬‬

‫‪zt − zt −1 = δ + at − θ1at , at‬‬

‫أي‬
‫‪٤٨‬‬

‫‪WN ( 0, σ 2 ) ,‬‬

‫‪θ1 < 1‬‬

‫‪zt = δ + zt −1 + at − θ1at , at‬‬

‫ﺛﺎﻟﺜﺎ‪ :‬ﻧﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ﺑﺈﻧﺠﺮاف ‪ Random Walk with Trend Model‬أو‬
‫)‪: ARIMA(٠،١،٠‬‬
‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬
‫) ‪WN ( 0, σ 2‬‬

‫‪φ0 ( B )(1 − B ) zt = δ + θ 0 ( B ) at , at‬‬

‫) ‪WN ( 0, σ 2‬‬

‫أي‬

‫)‬

‫‪2‬‬

‫‪WN ( 0, σ‬‬

‫‪at‬‬

‫‪(1 − B ) zt = δ + at ,‬‬

‫‪zt = δ + zt −1 + at , at‬‬

‫داﻟﺔ اﻷوزان ) ‪ ψ ( B‬وﺗﻤﺜﻴﻞ ﻧﻤﺎذج )‪:ARMA(p,q‬‬
‫ﺳﺒﻖ أن آﺘﺒﻨﺎ ﻧﻤﺎذج ) ‪ ARMA ( p, q‬ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪WN ( 0, σ 2‬‬

‫أو ﺑﺸﻜﻞ اﻹﻧﺤﺮاف ﻋﻦ اﻟﻤﺘﻮﺳﻂ‬

‫)‬

‫‪2‬‬

‫‪WN ( 0, σ‬‬

‫‪φ p ( B ) zt = δ + θ q ( B ) a t , a t‬‬
‫‪φ p ( B )( zt − µ ) = θ q ( B ) at , at‬‬

‫ﻓﻲ آﻠﺘﺎ اﻟﺤﺎﻟﺘﻴﻦ ﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻰ ﻋﺎﻣﻞ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ) ‪ φ p ( B‬ﻧﺠﺪ‬
‫) ‪WN ( 0, σ 2‬‬

‫) ‪θq ( B‬‬
‫‪a , at‬‬
‫‪φp (B) t‬‬

‫) ‪WN ( 0, σ 2‬‬

‫‪+‬‬

‫‪δ‬‬

‫)‪φ p (1‬‬

‫) ‪θq ( B‬‬
‫‪a , at‬‬
‫‪φp (B) t‬‬

‫= ‪zt‬‬

‫= ‪zt − µ‬‬

‫) ‪θq ( B‬‬
‫ﻻﺣﻆ ان ﻟﻠﻨﻤﺎذج اﻟﻤﺴﺘﻘﺮة اﻟﻨﺴﺒﺔ‬
‫)‪φp (B‬‬
‫‪δ‬‬
‫ﺗﻘﻊ ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة اﻳﻀﺎ ‪= µ‬‬
‫وﻟﻬﺬا ﺳﻮف ﻧﻜﺘﻔﻲ ﺑﺸﻜﻞ اﻹﻧﺤﺮاف ﻋﻦ اﻟﻤﺘﻮﺳﻂ ﻓﻲ‬
‫)‪φ p (1‬‬

‫ﺗﺸﻜﻞ ﻣﺴﻠﺴﻠﺔ ﻣﺘﻘﺎرﺑﺔ وذﻟﻚ ﻷن ﺟﺬور ‪φ p ( B ) = 0‬‬

‫ﻣﻨﺎﻗﺸﺘﻨﺎ اﻟﺘﺎﻟﻴﺔ‬
‫) ‪WN ( 0, σ 2‬‬

‫اﻟﻤﺴﻠﺴﻠﺔ اﻟﻤﺘﻘﺎرﺑﺔ‬

‫واﻟﺘﻲ ﺗﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫)‪θ (B‬‬
‫‪zt − µ = q‬‬
‫‪a , at‬‬
‫‪φp (B) t‬‬
‫) ‪θq ( B‬‬
‫)‪φp (B‬‬

‫)‪θq ( B‬‬
‫‪= ψ 0 B 0 + ψ 1 B1 + ψ 2 B 2 + ψ 3 B 3 + L , ψ 0 = 1‬‬
‫)‪φp (B‬‬

‫ﺗﺴﻤﻰ داﻟﺔ اﻷوزان‪.‬‬

‫‪٤٩‬‬

‫= )‪ψ (B‬‬

‫= )‪ψ (B‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١٧‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج ) ‪ ARMA ( p, q‬اﻟﻤﺴﺘﻘﺮ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬
‫)‪θq ( B‬‬
‫‪= ψ 0 B 0 + ψ 1 B1 + ψ 2 B 2 + ψ 3 B 3 + L , ψ 0 = 1‬‬
‫)‪φp (B‬‬

‫= )‪ψ (B‬‬

‫∞ )‪θq ( B‬‬
‫‪= ∑ψ B j , ψ 0 = 1‬‬
‫‪φ p ( B ) j =0 j‬‬

‫= )‪ψ (B‬‬

‫ﺣﻴﺚ اﻷوزان هﻲ ‪ψ 0 = 1,ψ 1 ,ψ 2 ,ψ 3 ,L‬‬
‫) ‪θq ( B‬‬
‫‪a , at‬‬
‫‪φp (B) t‬‬

‫ﻣﻼﺣﻈﺔ‪ :‬اﻟﻨﻤﻮذج اﻟﺬي ﻋﻠﻰ اﻟﺸﻜﻞ ) ‪WN ( 0, σ 2‬‬

‫= ‪ zt − µ‬ﻳﺴﻤﻰ ﺗﻤﺜﻴﻞ‬

‫اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻼﻧﻬﺎﺋﻲ ﻟﻨﻤﺎذج ) ‪. ARMA ( p, q‬‬

‫أﻣﺜﻠﺔ ﻟﺪاﻟﺔ اﻷوزان ﻟﺒﻌﺾ اﻟﻨﻤﺎذج‪:‬‬
‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪: AR(١‬‬
‫ﻧﻤﻮذج )‪ AR(١‬ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪φ1 ( B )( zt − µ ) = θ 0 ( B ) at , at WN ( 0, σ 2‬‬

‫‪(1 − φ1B )( zt − µ ) = at‬‬
‫‪1‬‬
‫‪a‬‬
‫‪(1 − φ1B ) t‬‬

‫= ‪zt − µ‬‬

‫‪zt − µ = ψ ( B ) at‬‬

‫ﺣﻴﺚ‬
‫‪1‬‬
‫) ‪(1 − φ1B‬‬

‫= )‪ψ (B‬‬

‫ﺳﻮف ﻧﻮﺟﺪ اﻷوزان ‪ ψ 1 ,ψ 2 ,ψ 3 ,L‬ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺘﺎﻟﻴﺔ‬
‫‪1‬‬
‫) ‪(1 − φ1B‬‬

‫= )‪ψ (B‬‬

‫‪ψ ( B )(1 − φ1B ) ≡ 1‬‬

‫‪B 2 + ψ 3 B 3 + L) (1 − φ1B ) ≡ 1‬‬

‫‪2‬‬

‫‪(1 + ψ B + ψ‬‬
‫‪1‬‬

‫ﻻﺣﻆ أن اﻟﻌﻼﻗﺔ اﻷﺧﻴﺮة هﻲ ﻋﻼﻗﺔ ﺗﻜﺎﻓﺆ أي ان ﻣﻌﺎﻣﻼت ‪ B j‬ﻋﻠﻰ ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﻣﺘﺴﺎوﻳﺔ‪.‬‬
‫وﺑﻤﺴﺎواة ﻣﻌﺎﻣﻼت ‪ B j‬ﻋﻠﻰ ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﻧﺠﺪ‬
‫‪φ1 < 1‬‬

‫‪+ ψ 3 B 3 + L) (1 − φ1B ) ≡ 1,‬‬

‫‪2‬‬

‫‪(1 + ψ B + ψ B‬‬
‫‪2‬‬

‫‪1‬‬

‫‪B 0 : (1)(1) ≡ 1‬‬
‫‪B1 : ψ 1 − φ1 ≡ 0 ⇒ ψ 1 = φ1‬‬
‫‪B 2 : ψ 2 − ψ 1φ1 ≡ 0 ⇒ ψ 2 = ψ 1φ1 = φ12‬‬
‫‪B 3 : ψ 3 − ψ 2φ1 ≡ 0 ⇒ ψ 3 = ψ 2φ1 = φ13‬‬
‫‪M‬‬
‫‪B j : ψ j − ψ j −1φ1 ≡ 0 ⇒ ψ j = ψ j −1φ1 = φ1j‬‬

‫أي ان اﻷوزان ﻟﻨﻤﻮذج )‪ AR(١‬هﻲ‬
‫‪٥٠‬‬

ψ j = φ1j , φ1 < 1

: MA(١) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬
‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬MA(١) ‫ﻧﻤﻮذج‬

φ0 ( B )( zt − µ ) = θ1 ( B ) at , at WN ( 0, σ 2 )

( zt − µ ) = (1 − θ1B ) at
zt − µ = ψ ( B ) at
‫ﺣﻴﺚ‬

ψ ( B ) = (1 − θ1B )

‫ ﻋﻠﻰ ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﻧﺠﺪ‬B j ‫ﺑﻤﺴﺎواة ﻣﻌﺎﻣﻼت‬

ψ 1 = −θ1 , ψ 2 = ψ 3 = L = 0

‫أي‬
j=0
j =1

⎧ 1,

ψ j = ⎨ −θ1 ,
⎪ 0,

j≥2

φ2 ( B )( zt − µ ) = θ 0 ( B ) at , at WN ( 0,σ 2 )

(1 − φ B − φ B ) ( z − µ ) = a

: AR(٢) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬
‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬AR(٢) ‫ﻧﻤﻮذج‬

2

1

zt − µ =

2

t

t

1
a
(1 − φ1B − φ2 B 2 ) t

zt − µ = ψ ( B ) at

‫ﺣﻴﺚ‬
ψ (B) =

1
(1 − φ1B − φ2 B 2 )

ψ ( B ) (1 − φ1B − φ2 B ) ≡ 1

‫ ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ‬ψ 1 ,ψ 2 ,ψ 3 ,L ‫و ﻧﻮﺟﺪ اﻷوزان‬

2

(1 + ψ B + ψ B
1

2

2

+ ψ 3 B 3 + L)(1 − φ1B − φ2 B 2 ) ≡ 1

B1 : ψ 1 − φ1 = 0 ⇒ ψ 1 = φ1
2
B 2 : ψ 2 − φψ
1 1 − φ2 = 0 ⇒ ψ 2 = φψ
1 1 + φ2 = φ1 + φ 2

B 3 : ψ 3 − φψ
1 2 − φ2ψ 1 = 0 ⇒ ψ 3 = φψ
1 2 + φ2ψ 1
M
B j : ψ j − φψ
1 j −1 − φ2ψ j − 2 = 0 ⇒ ψ j = φψ
1 j −1 + φ 2ψ j − 2

‫ هﻲ‬AR(٢) ‫أي ان اﻷوزان ﻟﻨﻤﻮذج‬

٥١

j=0

⎧ 1,
⎪φ ,
⎪ 1
ψj =⎨ 2
⎪ φ1 + φ2 ,
⎪⎩φψ
1 j −1 + φ2ψ j − 2 ,

j =1
j=2
j≥3

: MA(٢) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬
‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬MA(٢) ‫ﻧﻤﻮذج‬

φ0 ( B )( zt − µ ) = θ 2 ( B ) at , at WN ( 0,σ 2 )

( zt − µ ) = (1 − θ1B − θ 2 B 2 ) at
zt − µ = ψ ( B ) at
ψ ( B ) = (1 − θ1B − θ 2 B

2

‫ﺣﻴﺚ‬

)

‫ ﻋﻠﻰ ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﻧﺠﺪ‬B j ‫ﺑﻤﺴﺎواة ﻣﻌﺎﻣﻼت‬

ψ 1 = −θ1 , ψ 2 = −θ 2 , ψ 3 = ψ 4 = ψ 5 L = 0

‫أي‬
j=0
j =1
j=2
j≥2

⎧ 1,
⎪ −θ ,

ψj =⎨ 1
⎪ −θ 2 ,
⎪⎩ 0,

φ1 ( B )( zt − µ ) = θ1 ( B ) at , at

: ARMA(١،١) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬
‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬ARMA(١،١) ‫ﻧﻤﻮذج‬

WN ( 0, σ 2 )

(1 − φ1B )( zt − µ ) = (1 − θ1B ) at
(1 − θ1B ) a
zt − µ =
(1 − φ1B ) t
zt − µ = ψ ( B ) a t
ψ (B) =

(1 − θ1B )
(1 − φ1B )

ψ ( B )(1 − φ1B ) ≡ (1 − θ1B )

(1 + ψ B + ψ
1

‫ﺣﻴﺚ‬

2

‫ ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ‬ψ 1 ,ψ 2 ,ψ 3 ,L ‫و ﻧﻮﺟﺪ اﻷوزان‬

B 2 + ψ 3 B 3 + L) (1 − φ1B ) ≡ (1 − θ1B )

B1 : ψ 1 − φ1 = −θ1 ⇒ ψ 1 = φ1 − θ1
B 2 : ψ 2 − φψ
1 1 = 0 ⇒ ψ 2 = φψ
1 1 = φ1 (φ1 − θ1 )
2
B 3 : ψ 3 − φψ
1 2 = 0 ⇒ ψ 3 = φψ
1 2 = φ1 (φ1 − θ1 )

M

j −1
B j : ψ j − φψ
(φ1 − θ1 )
1 j −1 = 0 ⇒ ψ j = φψ
1 j −1 = φ1

‫ هﻲ‬ARMA(١،١) ‫أي ان اﻷوزان ﻟﻨﻤﻮذج‬
٥٢

j −1
ψ j = φψ
(φ1 − θ1 ) ,
1 j −1 = φ1

j ≥ 1,

φ1 < 1, φ1 ≠ θ1

: ARI(١) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬
‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬ARI(١) ‫ﻧﻤﻮذج‬

φ1 ( B )(1 − B )( zt − µ ) = at , at WN ( 0,σ 2 )
zt − µ =

1
a
(1 − φ1B )(1 − B ) t

zt − µ = ψ ( B ) at

‫ﺣﻴﺚ‬
ψ ( B) =

1
(1 − φ1B )(1 − B )

‫ ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ‬ψ 1 ,ψ 2 ,ψ 3 ,L ‫و ﻧﻮﺟﺪ اﻷوزان‬

ψ ( B )(1 − φ1B )(1 − B ) ≡ 1

(1 +ψ B +ψ
(1 +ψ B +ψ
1

2

1

2

B 2 + ψ 3 B 3 + L) (1 − φ1B )(1 − B ) ≡ 1

B 2 + ψ 3 B 3 + L) (1 − (φ1 + 1) B + φ1B 2 ) ≡ 1

B1 : ψ 1 − (φ1 + 1) = 0 ⇒ ψ 1 = φ1 + 1
B 2 : ψ 2 − (φ1 + 1)ψ 1 + φ1 = 0 ⇒ ψ 2 = (φ1 + 1)ψ 1 + φ1 = (φ1 + 1) + φ1
2

B 3 : ψ 3 − (φ1 + 1)ψ 2 + φψ
1 1 = 0 ⇒ ψ 3 = (φ1 + 1)ψ 2 − φψ
1 1
M

B j : ψ j − (φ1 + 1)ψ j −1 + φψ
1 j −2
1 j − 2 = 0 ⇒ ψ j = (φ1 + 1)ψ j −1 − φψ

‫ هﻲ‬ARI(١) ‫أي ان اﻷوزان ﻟﻨﻤﻮذج‬
j=0
j =1

⎧ 1,
⎪ φ + 1,
⎪ 1
ψj =⎨
2
⎪ (φ1 + 1) + φ1 ,
⎪(φ1 + 1)ψ j −1 − φψ
1 j −2 ,

j=2
j≥3

‫وأﺧﻴﺮا ﻧﻮﺟﺪ‬
ARIMA(١،٠،١) ‫ أو‬Random Walk Mdel ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ‬
‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬
zt = zt −1 + at , at

WN ( 0, σ 2 )

zt − zt −1 = at , at

WN ( 0, σ

2

‫أي‬

)

(1 − B ) zt = at
zt =

1
a
(1 − B ) t

‫ﺣﻴﺚ‬
1
ψ ( B) =
(1 − B )

٥٣

‫و ﻧﻮﺟﺪ اﻷوزان ‪ ψ 1 ,ψ 2 ,ψ 3 ,L‬ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ‬

‫‪ψ ( B )(1 − B ) ≡ 1‬‬
‫‪+ ψ 3 B 3 + L) (1 − B ) ≡ 1‬‬

‫‪2‬‬

‫‪(1 + ψ B + ψ B‬‬
‫‪1‬‬

‫‪2‬‬

‫‪B1 : ψ 1 − 1 = 0 ⇒ ψ 1 = 1‬‬
‫‪B 2 :ψ 2 −ψ 1 = 0 ⇒ ψ 2 = ψ 1 = 1‬‬
‫‪B 3 :ψ 3 −ψ 2 = 0 ⇒ ψ 3 = ψ 2 = 1‬‬
‫‪M‬‬
‫‪B j : ψ j − ψ j −1 = 0 ⇒ ψ j = ψ j −1 = 1‬‬

‫أي ان اﻷوزان ﻟﻨﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ) ‪ ARIMA ( 0,1,0‬هﻲ‬
‫‪ψ j = 1,‬‬

‫‪j ≥1‬‬

‫ﺑﻌﺾ ﺧﻮاص داﻟﺔ اﻷوزان ) ‪: ψ ( B‬‬
‫ﺳﺒﻖ أن آﺘﺒﻨﺎ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪ ARMA(p,q‬ﻋﻠﻰ اﻟﺸﻜﻞ‬
‫) ‪WN ( 0, σ 2‬‬

‫وﺑﻜﺘﺎﺑﺔ هﺬﻩ اﻟﻌﻼﻗﻪ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫‪zt − µ = ψ ( B ) a t , a t‬‬

‫‪zt − µ = at + ψ 1at −1 + ψ 2 at −2 + ψ 3at −3 + L‬‬
‫∞‬

‫‪= ∑ψ j at − j , ψ 0 = 1‬‬
‫‪j =0‬‬

‫وإذا اﻓﺘﺮﺿﻨﺎ ان اﻷوزان ﺗﺘﻘﺎرب اي ∞ < ‪ψ 2j‬‬

‫∞‬

‫‪j =0‬‬

‫∑‬

‫ﻓﺈﻧﻪ ﻳﻤﻜﻦ إﺛﺒﺎت اﻟﻨﻈﺮﻳﺔ اﻟﺘﺎﻟﻴﺔ‪:‬‬

‫ﻧﻈﺮﻳﺔ ‪:١‬‬
‫ﻟﻨﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪ ARMA(p,q‬اﻟﻤﺴﺘﻘﺮ واﻟﺬي ﻳﻜﺘﺐ‬
‫ﻋﻠﻰ اﻟﺸﻜﻞ‬
‫∞ < ‪WN ( 0, σ 2 ) , ψ 0 = 1, ∑ j =0ψ 2j‬‬
‫∞‬

‫∞‬

‫‪zt − µ =∑ψ j at − j , at‬‬
‫‪j =0‬‬

‫‪ -١‬اﻟﻤﺘﻮﺳﻂ هﻮ‬

‫‪E ( zt ) = µ , ∀t‬‬

‫‪ -٢‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬
‫∞‬

‫‪j +k‬‬

‫‪, k = 0,1, 2,L‬‬

‫‪∑ψ ψ‬‬
‫‪j‬‬

‫‪j =0‬‬

‫‪2‬‬
‫‪j‬‬

‫∞‬

‫‪∑ψ‬‬

‫= ‪ρk‬‬

‫‪j=0‬‬

‫ﻣﺜﺎل‪ :‬ﻟﻨﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ )‪ AR(١‬وﺟﺪﻧﺎ ﺳﺎﺑﻘﺎ داﻟﺔ اﻷوزان‬
‫‪ψ j = φ , φ1 < 1‬‬
‫‪j‬‬
‫‪1‬‬

‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬

‫‪٥٤‬‬

‫‪= φ1k , k = 0,1, 2,L‬‬

‫‪φ1k‬‬
‫‪1 − φ12‬‬
‫‪1‬‬
‫‪1 − φ12‬‬

‫∞‬

‫=‬

‫‪∑φ1jφ1j+k‬‬
‫‪j =0‬‬
‫∞‬

‫‪∑φ‬‬

‫‪2j‬‬
‫‪1‬‬

‫‪j =0‬‬

‫∞‬

‫=‬

‫‪∑ψ jψ j+k‬‬
‫‪j =0‬‬

‫‪2‬‬
‫‪j‬‬

‫‪∑ψ‬‬

‫وهﻲ ﻧﻔﺲ اﻟﻨﺘﻴﺠﺔ اﻟﺴﺎﺑﻘﺔ‬
‫ﺗﻤﺮﻳﻦ‪ :‬ﺑﺄﺳﺘﺨﺪام ﻧﻈﺮﻳﺔ ‪ (٢) ٢‬أوﺟﺪ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻨﻤﺎذج اﻟﺘﺎﻟﻴﺔ‬
‫)‪AR(٢), MA(١), MA(٢), ARMA(١،١), ARMA(٢،١), ARMA(١،٢‬‬

‫‪٥٥‬‬

‫∞‬

‫‪j=0‬‬

‫= ‪ρk‬‬

‫اﻟﻔﺼﻞ اﻟﺮاﺑﻊ‬
‫اﻟﺘﻨﺒﺆات ذات ﻣﺘﻮﺳﻂ ﻣﺮﺑﻊ اﻟﺨﻄﺄ اﻷدﻧﻰ ﻟﻨﻤﺎذج )‪ARMA(p,q‬‬
‫‪Minimum Mean Square Error Forecasts for‬‬
‫‪ARMA(p,q) Models‬‬
‫ﻓﻲ ﺍﻟﻔﻘﺭﺓ ﺍﻟﺴﺎﺒﻘﺔ ﻜﺘﺒﻨﺎ ﻨﻤﻭﺫﺝ ﺍﻹﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ‪-‬ﺍﻟﻤﺘﻭﺴﻁ ﺍﻟﻤﺘﺤﺭﻙ ﻤﻥ ﺍﻟﺩﺭﺠﺔ )‪ARMA(p,q‬‬
‫ﺍﻟﻤﺴﺘﻘﺭ ﻋﻠﻰ ﺍﻟﺸﻜل‬
‫∞ < ‪WN ( 0, σ 2 ) , ψ 0 = 1, ∑ j =0ψ 2j‬‬
‫∞‬

‫∞‬

‫‪zt − µ =∑ψ j at − j , at‬‬
‫‪j =0‬‬

‫ﺃﻭ‬

‫‪zt − µ = at + ψ 1at −1 + ψ 2at −2 + ψ 3at −3 +L‬‬
‫∞‬

‫‪=∑ψ j at − j , ψ 0 = 1‬‬
‫‪j =0‬‬

‫ﻤﻼﺤﻅﺔ‪ :‬ﻫﺫﺍ ﻴﻨﻁﺒﻕ ﺃﻴﻀﺎ ﻋﻠﻰ ﻨﻤﺎﺫﺝ )‪ ARIMA(p,d,q‬ﺒﺸﻜل ﻋﺎﻡ‪.‬‬
‫ﻟﻤﺘﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﻤﺸﺎﻫﺩﺓ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺍﻟﺘﻨﺒﺅﺍﺕ ‪ zn ( l ) , l ≥ 1‬ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬
‫‪ zn +l , l ≥ 1‬ﻴﻤﻜﻥ ﺍﻥ ﺘﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬
‫‪zn ( l ) = ξ 0an + ξ1an −1 + ξ 2an −2 + L, l ≥ 1‬‬

‫ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +l , l ≥ 1‬ﺘﻜﺘﺏ ﺒﺩﻻﻟﺔ ﺍﻟﻨﻤﻭﺫﺝ ﻜﺎﻟﺘﺎﻟﻲ‬

‫‪zn + l − µ = an +l + ψ 1an + l−1 + L + ψ l−1an +1 + ψ l an + ψ l+1an −1 + L, l ≥ 1‬‬

‫ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻟﺨﻁﺄ ﻴﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ )ﺃﻨﻅﺭ ﺘﻌﺭﻴﻑ ‪( ٥‬‬

‫‪2‬‬

‫⎦⎤‪E ⎡⎣ zn + l − zn ( l )⎤⎦ = E ⎡⎣ an + l + ψ 1an + l−1 + L + ψ l−1an +1 + (ψ l − ξ 0 ) an + (ψ l+1 − ξ1 ) an −1 + L‬‬
‫‪2‬‬

‫∞‬

‫‪= (1 + ψ 12 + L + ψ l2−1 ) σ 2 + ∑ (ψ l+ j − ξ j ) σ 2‬‬
‫‪2‬‬

‫‪j =0‬‬

‫ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻟﺨﻁﺄ ﺍﻷﺩﻨﻰ ﻴﻨﺘﺞ ﻤﻥ ﺘﺼﻐﻴﺭ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺴﺎﺒﻘﺔ ﺒﺎﻟﻨﺴﺒﺔ ﻟﻸﻭﺯﺍﻥ ‪ ξ j‬ﻟﺠﻤﻴﻊ ﻗﻴﻡ ‪j‬‬
‫ﻭﻫﺫﺍ ﻴﻤﻜﻥ ﺇﺫﺍ ﻭﻓﻘﻁ ﺇﺫﺍ ﺤﻘﻘﺕ ﺍﻷﻭﺯﺍﻥ ‪ ξ j‬ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‬
‫‪j = 0,1, 2,L , l ≥ 1‬‬

‫‪ξ j = ψ l+ j ,‬‬

‫ﻭﻋﻠﻴﻪ ﻓﺈﻥ ﺍﻟﺘﻨﺒﺅﺍﺕ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻟﺨﻁﺄ ﺍﻷﺩﻨﻰ ‪ MMSE Forecasts‬ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬
‫‪zn ( l ) = ψ l an + ψ l+1an −1 + ψ l+2an −2 + L, l ≥ 1‬‬

‫ﻨﻅﺭﻴﺔ ‪:٢‬‬
‫ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‪:‬‬

‫‪en ( l ) = zn +l − zn ( l ) = an +l + ψ 1an +l−1 + ψ 2an +l−2 + L + ψ l−1an+1 , l ≥ 1‬‬

‫ﻭﺘﺒﺎﻴﻥ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‪:‬‬
‫‪V ⎡⎣ en ( l )⎤⎦ = σ 2 (1 + ψ 12 + ψ 22 + L + ψ l2−1 ) , l ≥ 1‬‬

‫‪٥٦‬‬

‫ﺍﻟﺼﻴﻐﺔ ‪ zn ( l ) = ψ lan + ψ l+1an−1 + ψ l+2an −2 + L, l ≥ 1‬ﻏﻴﺭ ﻋﻤﻠﻴﺔ ﻹﻴﺠﺎﺩ ﺍﻟﺘﻨﺒﺅﺍﺕ ﻟﻠﻘﻴﻡ‬
‫ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn+l , l ≥ 1‬ﻭﺫﻟﻙ ﻷﻨﻨﺎ ﻨﺤﺘﺎﺝ ﺇﻟﻰ ﻤﻌﺭﻓﺔ ﺍﻟﻘﻴﻡ } ‪. {a1 , a2 ,L, an−1 , an‬‬

‫ﺘﻌﺭﻴﻑ ‪ : ١٨‬ﻤﺠﻤﻭﻋﺔ ﺍﻟﻤﻌﻠﻭﻤﺎﺕ )} ‪ I ({z1 , z2 ,L , zn −1 , zn‬ﺘﻜﺎﻓﺊ ﻤﺠﻤﻭﻋﺔ ﺍﻟﻤﻌﻠﻭﻤﺎﺕ‬
‫)} ‪ I ({a1 , a2 ,L , an −1 , an‬ﻭﺫﻟﻙ ﺒﺎﻟﻤﻌﻨﻰ ﺃﻥ ﺍﻟﻤﺠﻤﻭﻋﺔ } ‪ {a1 , a2 ,L, an−1 , an‬ﺘﺤﺘﻭﻯ ﻋﻠﻰ ﻨﻔﺱ‬
‫ﺍﻟﻤﻌﻠﻭﻤﺎﺕ ﻋﻥ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪. {z1 , z2 ,L, zn −1 , zn‬‬
‫ﻤﻼﺤﻅﺔ‪ :‬ﺍﻟﻤﺘﺴﻠﺴﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﻴﻤﻜﻥ ﻤﺸﺎﻫﺩﺘﻬﺎ ﻭﻗﻴﺎﺴﻬﺎ ﻭﻟﻜﻥ ﺍﻟﻤﺘﻠﺴﻠﺔ‬
‫} ‪ {a1 , a2 ,L, an−1 , an‬ﻻﻴﻤﻜﻥ ﻤﺸﺎﻫﺩﺘﻬﺎ ﺃﻭ ﻗﻴﺎﺴﻬﺎ‪.‬‬

‫ﻨﻅﺭﻴﺔ ‪ : ٣‬ﺍﻟﻤﺘﻨﺒﺊ ﺫﺍ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻟﺨﻁﺄ ﺍﻷﺩﻨﻰ ‪ MMSE Forecasts‬ﻴﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬
‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬
‫ﺃﻱ ﻫﻭ ﺍﻟﺘﻭﻗﻊ ﺍﻟﺸﺭﻁﻲ ﻟﻠﻘﻴﻤﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +l , l ≥ 1‬ﻤﻌﻁﻰ } ‪. {z1 , z2 ,L, zn −1 , zn‬‬
‫ﺘﺴﺘﺨﺩﻡ ﻨﻅﺭﻴﺔ ‪ ٢‬ﻋﻤﻠﻴﺎ ﻹﻴﺠﺎﺩ ﻗﻴﻡ ﺍﻟﺘﻨﺒﺅﺍﺕ ﺒﺩﻻ ﻤﻥ ﺍﻟﺼﻴﻐﺔ‬
‫‪zn ( l ) = ψ l an + ψ l+1an −1 + ψ l+2an −2 + L, l ≥ 1‬‬

‫ﻭﺫﻟﻙ ﺘﺒﻌﺎ ﻟﻠﻤﻼﺤﻅﺔ ﺍﻟﺴﺎﺒﻘﺔ‪.‬‬

‫ﻗﺎﻋﺩﺓ ‪:٢‬‬
‫‪⎧a , j ≤ 0‬‬
‫‪1 − E ( an + j zn , zn −1 ,L) = ⎨ n + j‬‬
‫‪j>0‬‬
‫‪⎩ 0,‬‬
‫‪j≤0‬‬
‫‪⎧ zn + j ,‬‬
‫⎨ = )‪2 − E ( zn + j zn , zn −1 ,L‬‬
‫‪⎩ zn ( j ) , j > 0‬‬

‫ﻨﻅﺭﻴﺔ ‪ ٣‬ﻤﻊ ﺍﻟﻘﺎﻋﺩﺓ ‪ ٢‬ﺘﻌﻁﻲ ﻁﺭﻴﻘﺔ ﻋﻤﻠﻴﺔ ﻭﺴﻬﻠﺔ ﻹﻴﺠﺎﺩ ﺘﻨﺒﺅﺍﺕ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬
‫‪zn + l , l ≥ 1‬‬

‫ﺍﻟﺩﺍﻟﺔ ‪ zn ( l ) , l ≥ 1‬ﻜﺩﺍﻟﺔ ﻟﺯﻤﻥ ﺍﻟﺘﻘﺩﻡ‬

‫‪l ≥1‬‬

‫ﺘﻌﺭﻴﻑ ‪:١٩‬‬
‫ﻋﻨﺩ ﻨﻘﻁﺔ ﺍﻻﺼل ﻟﻠﺯﻤﻥ ‪ n‬ﺘﺴﻤﻰ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ‪.‬‬

‫دوال اﻟﺘﻨﺒﺆ ﻟﻨﻤﺎذج )‪: ARIMA(p,d,q‬‬
‫ﺍﻭﻻ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: AR(١‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬
‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ AR(١‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬

‫‪٥٧‬‬

‫) ∞ ‪WN ( 0, σ 2 ) , φ1 < 1, µ ∈ ( −∞,‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + at , at‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn +l , l ≥ 1‬‬
‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬
‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ +E ⎡⎣ ⎡⎣φ1 ( zn + l−1 − µ ) + an +l ⎤⎦ zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫‪=µ +E ⎡⎣φ1 ( zn + l−1 − µ ) zn , zn −1 ,L + an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫‪= µ +φ1E ⎡⎣( zn + l−1 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫ﺃﻱ‬
‫‪zn ( l ) = µ +φ1E ⎣⎡ ( zn + l−1 zn , zn −1 ,L) − µ ⎦⎤ + E ⎡⎣ an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ ‪٢‬‬

‫⎦⎤‪l = 1: zn (1) = µ +φ1E ⎣⎡ ( zn zn , zn −1 ,L) − µ ⎦⎤ + E ⎡⎣ an +1 zn , zn −1 ,L‬‬
‫) ‪= µ +φ1 ( zn − µ‬‬

‫⎦⎤‪l = 2 : zn ( 2 ) = µ +φ1E ⎡⎣( zn +1 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + 2 zn , zn −1 ,L‬‬
‫⎦⎤ ‪= µ +φ1 ⎡⎣ zn (1) − µ‬‬

‫⎦⎤‪l = 3 : zn ( 3) = µ +φ1E ⎡⎣( zn + 2 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an +3 zn , zn −1 ,L‬‬
‫⎦⎤ ‪= µ +φ1 ⎡⎣ zn ( 2 ) − µ‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = µ +φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ , l ≥ 1‬‬

‫ﻭﻫﻲ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪AR(١‬‬

‫ﺘﻌﺭﻴﻑ ‪:٢٠‬‬
‫ﺸﺭﻁ ﺍﻹﺴﺘﻤﺭﺍﺭ ‪ Continuity Condition‬ﻴﺘﻁﻠﺏ ﺃﻨﻪ ﻋﻨﺩﻤﺎ ﺘﻜﻭﻥ ‪ l = 1‬ﻓﺈﻥ‬
‫‪zn ( l − 1) = zn ( 0 ) = zn‬‬

‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٢‬ﺘﺒﺎﻴﻥ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﺘﻌﻁﻰ ﻤﻥ ﺍﻟﻌﻼﻗﺔ‬
‫‪V ⎡⎣ en ( l )⎤⎦ = σ 2 (1 + ψ 12 + ψ 22 + L + ψ l2−1 ) , l ≥ 1‬‬

‫ﺴﺒﻕ ﺃﻥ ﺍﺸﺘﻘﻘﻨﺎ ﺩﺍﻟﺔ ﺍﻷﻭﺯﺍﻥ ﻟﻨﻤﻭﺫﺝ )‪ AR(١‬ﻭﻫﻲ‬
‫‪ψ j = φ1j , φ1 < 1‬‬

‫‪), l ≥1‬‬

‫ﻭﺒﺎﻟﺘﻌﻭﻴﺽ ﻓﻲ ﺼﻴﻐﺔ ﺍﻟﺘﺒﺎﻴﻥ ﻨﺠﺩ‬

‫(‬

‫( ‪V ⎡⎣ en ( l )⎤⎦ = σ 2 1 + φ12 + φ14 + L + φ1‬‬

‫)‪2 l −1‬‬

‫‪1 − φ12 l‬‬
‫‪, l ≥1‬‬
‫‪1 − φ12‬‬

‫‪=σ2‬‬

‫ﻤﺜﺎل‪ :‬ﻤﺘﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﻤﺸﺎﻫﺩﻩ ﻭﺠﺩ ﺍﻨﻬﺎ ﺘﺘﺒﻊ ﺍﻟﻨﻤﻭﺫﺝ‬
‫‪٥٨‬‬

‫) ‪WN ( 0,0.024‬‬

‫‪zt − 0.97 = 0.85 ( zt −1 − 0.97 ) + at , at‬‬

‫ﺇﺫﺍ ﻜﺎﻨﺕ ﺍﻟﻤﺸﺎﻫﺩﺓ ﺍﻷﺨﻴﺭﺓ ﻫﻲ ‪ ، z156 = 0.49‬ﺃﻭﺠﺩ ﺘﻨﺒﺅﺍﺕ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪z157 , z158 , z159‬‬

‫ﻭﺃﻭﺠﺩ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻟﻬﺎ‪.‬‬
‫ﺍﻟﺤل‪ :‬ﻤﻥ ﺍﻟﺼﻴﻐﺔ ‪ zn ( l ) = µ +φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ , l ≥ 1‬ﻨﺠﺩ‬
‫) ‪z156 (1) = 0.97+0.85 ( z156 − 0.97‬‬

‫‪= 0.97+0.85 ( 0.49 − 0.97 ) = 0.56‬‬
‫) ‪z156 ( 2 ) = 0.97+0.85 ( z156 (1) − 0.97‬‬
‫‪= 0.97+0.85 ( 0.56 − 0.97 ) = 0.62‬‬
‫) ‪z156 ( 3) = 0.97+0.85 ( z156 ( 2 ) − 0.97‬‬
‫‪= 0.97+0.85 ( 0.62 − 0.97 ) = 0.68‬‬

‫ﻭﺍﻟﺘﺒﺎﻴﻨﺎﺕ‬
‫‪1 − φ12 l‬‬
‫‪, l ≥1‬‬
‫‪1 − φ12‬‬

‫‪V ⎡⎣ en ( l )⎤⎦ = σ 2‬‬

‫‪V ⎡⎣ e156 (1)⎤⎦ = 0.024‬‬
‫‪= 0.041‬‬
‫‪= 0.054‬‬

‫‪4‬‬

‫)‪1 − ( 0.85‬‬

‫‪2‬‬

‫)‪1 − ( 0.85‬‬

‫‪6‬‬

‫)‪1 − ( 0.85‬‬

‫‪2‬‬

‫)‪1 − ( 0.85‬‬

‫‪V ⎡⎣ e156 ( 2 )⎤⎦ = 0.024‬‬
‫‪V ⎡⎣ e156 ( 2 )⎤⎦ = 0.024‬‬

‫ﺜﺎﻨﻴﺎ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: AR(٢‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬
‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ AR(٢‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬
‫‪WN ( 0, σ 2 ) , µ ∈ ( −∞, ∞ ) ,‬‬

‫‪zt = µ + φ1 ( zt −1 − µ ) + φ2 ( zt −2 − µ ) + at , at‬‬

‫‪φ2 − φ1 < 1, φ2 + φ1 < 1, φ2 < 1‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn +l , l ≥ 1‬‬
‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬
‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ +E ⎡⎣ ⎡⎣φ1 ( zn + l−1 − µ ) + φ2 ( zn + l−2 − µ ) + an + l ⎤⎦ zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫‪=µ +E ⎡⎣φ1 ( zn + l−1 − µ ) zn , zn −1 ,L + φ2 ( zn + l−2 − µ ) zn , zn −1 ,L + an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫‪= µ +φ1E ⎡⎣( zn +l−1 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣( zn + l−2 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫ﺃﻱ‬

‫‪٥٩‬‬

zn ( l ) = µ +φ1E ⎡⎣ ( zn + l −1 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣ ( zn + l −2 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + l zn , zn −1 ,L⎤⎦ , l ≥ 1

٢ ‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ‬

l = 1: zn (1) = µ +φ1E ⎡⎣( zn zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣( zn −1 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an +1 zn , zn −1 ,L⎤⎦
= µ +φ1 ( zn − µ ) + φ2 ( zn −1 − µ )

l = 2 : zn ( 2 ) = µ +φ1E ⎡⎣( zn +1 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣ ( zn zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an +2 zn , zn −1 ,L⎤⎦
= µ +φ1 ⎡⎣ zn (1) − µ ⎤⎦ + φ2 ( zn − µ )

l = 3 : zn ( 3) = µ +φ1E ⎡⎣( zn + 2 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣( zn +1 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an +3 zn , zn −1 ,L⎤⎦
= µ +φ1 ⎡⎣ zn ( 2 ) − µ ⎤⎦ +φ2 ⎡⎣ zn (1) − µ ⎤⎦

l = 4 : zn ( 4 ) = µ +φ1E ⎡⎣( zn +3 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣( zn + 2 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + 4 zn , zn −1 ,L⎤⎦
= µ +φ1 ⎡⎣ zn ( 3) − µ ⎤⎦ +φ2 ⎡⎣ zn ( 2 ) − µ ⎤⎦

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

zn ( l ) = µ +φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ + φ2 ⎡⎣ zn ( l − 2 ) − µ ⎤⎦ , l ≥ 1

AR(٢) ‫ﻭﻫﻲ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ‬
‫ ﻭﺩﺍﻟﺔ ﺍﻷﻭﺯﺍﻥ‬٢ ‫ﻭﻴﻤﻜﻥ ﺤﺴﺎﺏ ﺘﺒﺎﻴﻨﺎﺕ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻤﻥ ﻨﻅﺭﻴﺔ‬
⎧ 1,
⎪φ ,
⎪ 1
ψj =⎨ 2
⎪ φ1 + φ2 ,
⎪⎩φψ
1 j −1 + φ2ψ j − 2 ,

j=0
j =1
j=2
j≥3

: ARIMA(٠،١،١) ‫ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ‬:‫ﺜﺎﻟﺜﺎ‬

‫ ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬n ‫{ ﺤﺘﻰ ﺍﻟﺯﻤﻥ‬z1 , z2 ,L, zn −1 , zn } ‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ‬
‫ ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬ARIMA(٠،١،١) ‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ‬
zt = zt −1 + at − θ1at −1 , at

WN ( 0, σ 2 )

. zn+l , l ≥ 1 ‫ ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ‬zn +1 , zn +2 , zn+3 ,L ‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬
‫ ﻨﺠﺩ‬٣ ‫ﻤﻥ ﻨﻅﺭﻴﺔ‬
zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1

=E ( zn +l−1 zn , zn −1 ,L) + E ( an +l zn , zn −1 ,L) − θ1E ( an +l−1 zn , zn −1 ,L) , l ≥ 1

٢ ‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ‬

٦٠

‫‪zn ( l ) =E ( zn + l−1 zn , zn −1 ,L) + E ( an + l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫)‪l = 1: zn (1) =E ( zn zn , zn −1 ,L) + E ( an +1 zn , zn −1 ,L) − θ1E ( an zn , zn −1 ,L‬‬
‫‪= zn − θ1an‬‬

‫)‪l = 2 : zn ( 2 ) =E ( zn +1 zn , zn −1 ,L) + E ( an + 2 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L‬‬
‫)‪= zn (1‬‬
‫)‪l = 3 : zn ( 3) =E ( zn + 2 zn , zn −1 ,L) + E ( an +3 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L‬‬
‫) ‪= zn ( 2‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = zn ( l − 1) , l ≥ 2‬‬

‫ﻭﻫﻜﺫﺍ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪ ARIMA(٠،١،١‬ﺘﻌﻁﻰ‬
‫ﺒﺎﻟﻌﻼﻗﺔ‬
‫‪⎧ zn − θ1an , l = 1‬‬
‫⎨ = ) ‪zn ( l‬‬
‫‪⎩ zn ( l − 1) , l > 1‬‬

‫ﺭﺍﺒﻌﺎ ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: MA(١‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬
‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ MA(١‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬
‫) ‪WN ( 0, σ 2‬‬

‫‪zt = µ + at − θ1at −1 , at‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn +l , l ≥ 1‬‬
‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬
‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ + E ( an + l zn , zn −1 ,L) − θ1E ( an +l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ ‪٢‬‬

‫‪zn ( l ) =µ + E ( an + l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫)‪l = 1: zn (1) =µ + E ( an +1 zn , zn −1 ,L) − θ1E ( an zn , zn −1 ,L‬‬
‫‪= µ − θ1an‬‬

‫)‪l = 2 : zn ( 2 ) =µ + E ( an + 2 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L‬‬
‫‪=µ‬‬
‫)‪l = 3 : zn ( 3) =µ + E ( an +3 zn , zn −1 ,L) − θ1E ( an + 2 zn , zn −1 ,L‬‬
‫‪=µ‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = µ , l ≥ 2‬‬

‫ﻭﻫﻜﺫﺍ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪ MA(١‬ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬
‫‪⎧ µ − θ1an , l = 1‬‬
‫⎨ = ) ‪zn ( l‬‬
‫‪l≥2‬‬
‫‪⎩ µ,‬‬

‫ﺨﺎﻤﺴﺎ ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: MA(٢‬‬

‫‪٦١‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬
‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ MA(٢‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬
‫) ‪WN ( 0, σ 2‬‬

‫‪zt = µ + at − θ1at −1 − θ 2 at − 2 , at‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn+l , l ≥ 1‬‬
‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬
‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ + E ( an + l zn , zn −1 ,L) − θ1E ( an +l−1 zn , zn −1 ,L) − θ 2 E ( an +l−2 zn , zn −1 ,L) , l ≥ 1‬‬

‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ ‪٢‬‬

‫‪zn ( l ) =µ + E ( an + l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) − θ 2 E ( an + l−2 zn , zn −1 ,L) , l ≥ 1‬‬

‫)‪l = 1: zn (1) =µ + E ( an +1 zn , zn −1 ,L) − θ1E ( an zn , zn −1 ,L) − θ 2 E ( an −1 zn , zn −1 ,L‬‬
‫‪= µ − θ1an − θ 2an −1‬‬

‫)‪l = 2 : zn ( 2 ) =µ + E ( an + 2 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L) − θ 2 E ( an zn , zn −1 ,L‬‬
‫‪= µ − θ 2 an‬‬

‫)‪l = 3 : zn ( 3) =µ + E ( an +3 zn , zn −1 ,L) − θ1E ( an + 2 zn , zn −1 ,L) − θ 2 E ( an +1 zn , zn −1 ,L‬‬
‫‪=µ‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = µ , l ≥ 3‬‬

‫ﻭﻫﻜﺫﺍ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪ MA(٢‬ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬
‫‪⎧ µ − θ1an − θ 2an −1 , l = 1‬‬
‫⎪‬
‫‪zn ( l ) = ⎨ µ − θ 2 an ,‬‬
‫‪l=2‬‬
‫‪⎪ µ,‬‬
‫‪l≥3‬‬
‫⎩‬

‫ﺴﺎﺩﺴﺎ ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: ARMA(١،١‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬
‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ ARMA(١،١‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬
‫‪WN ( 0, σ 2 ) , φ1 ≠ θ1 , φ1 < 1‬‬

‫‪zt = µ + φ1 ( zt −1 − µ ) + at − θ1at −1 , at‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn +l , l ≥ 1‬‬
‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬
‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ + φ1E ⎡⎣( zn + l−1 − µ ) zn , zn −1 ,L⎤⎦ + E ( an + l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ ‪٢‬‬

‫‪٦٢‬‬

‫‪zn ( l ) =µ + φ1 E ⎡⎣( zn + l−1 − µ ) zn , zn −1 ,L⎤⎦ + E ( an +l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫)‪l = 1: zn (1) =µ + φ1E ⎣⎡( zn − µ ) zn , zn −1 ,L⎦⎤ + E ( an +1 zn , zn −1 ,L) − θ1E ( an zn , zn −1 ,L‬‬
‫‪= µ + φ1 ( zn − µ ) − θ1an‬‬

‫)‪l = 2 : zn ( 2 ) =µ + φ1 E ⎡⎣( zn +1 − µ ) zn , zn −1 ,L⎤⎦ + E ( an +2 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L‬‬
‫⎦⎤ ‪= µ + φ1 ⎡⎣ zn (1) − µ‬‬

‫)‪l = 3 : zn ( 3) =µ + φ1 E ⎡⎣( zn + 2 − µ ) zn , zn −1 ,L⎤⎦ + E ( an +3 zn , zn −1 ,L) − θ1E ( an + 2 zn , zn −1 ,L‬‬
‫⎦⎤ ‪= µ + φ1 ⎡⎣ zn ( 2 ) − µ‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = µ + φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ , l ≥ 2‬‬

‫ﻭﻫﻜﺫﺍ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪ ARMA(١،١‬ﺘﻌﻁﻰ‬
‫ﺒﺎﻟﻌﻼﻗﺔ‬
‫‪l =1‬‬
‫‪l≥2‬‬

‫‪⎧⎪ µ + φ1 ( zn − µ ) − θ1an ,‬‬
‫⎨ = ) ‪zn ( l‬‬
‫‪⎪⎩ µ + φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ ,‬‬

‫ﺘﻤﺭﻴﻥ‪:‬‬
‫ﻟﻨﻤﻭﺫﺝ )‪ ARMA(١،١‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬

‫‪WN ( 0, σ 2 ) , φ1 ≠ θ1 , φ1 < 1‬‬

‫‪zt = µ + φ1 ( zt −1 − µ ) + at − θ1at −1 , at‬‬

‫ﺒﺭﻫﻥ ﺍﻥ ﻋﻨﺩﻤﺎ ﺘﺅﻭل ‪ φ1 → 1‬ﻓﺈﻥ )‪ ARMA(1,1) → IMA(1,1‬ﻭﻤﻥ ﺜﻡ ﺃﻭﺠﺩ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ‬
‫ﻟﻨﻤﻭﺫﺝ )‪. IMA(١،١‬‬
‫ﺘﻤﺭﻴﻥ‪:‬‬
‫ﺃﻭﺠﺩ ﺩﻭﺍل ﺍﻟﺘﻨﺒﺅ ﻭﺘﺒﺎﻴﻥ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻟﻜل ﻤﻥ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪ARIMA(١،١،١), ARIMA(٢،١،٠), ARIMA(٠،١،٢), ARIMA(١،٢،٠),‬‬
‫‪ARIMA(٠،٢،١), ARIMA(٠،٢،٠).‬‬
‫ﺤﺩﻭﺩ ﺍﻟﺘﻨﺒﺅ ‪: Forecasting Limits‬‬
‫ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ‪ zn ( l ) , l ≥ 1‬ﻋﻨﺩ ﻗﻴﻤﺔ ﻤﻌﻴﻨﺔ ﺘﻌﻁﻲ ﻤﺎﻴﺴﻤﻰ ﺒﺘﻨﺒﺅ ﺍﻟﻨﻘﻁﺔ ‪ Point Forecast‬ﻭﺍﻟﺫﻱ‬
‫ﻻﻴﻜﻔﻲ ﺍﻭ ﻴﻔﻴﺩ ﻓﻲ ﺇﺘﺨﺎﺫ ﻗﺭﺍﺭﺍﺕ ﺇﺤﺼﺎﺌﻴﺔ ﻋﻥ ﺍﻟﻅﺎﻫﺭﺓ ﺍﻟﻌﺸﻭﺍﺌﻴﺔ ﺍﻟﻤﺩﺭﻭﺴﺔ ﻷﻥ‬
‫‪P ( Z n + m = zn ( m ) ) = 0, for some m > 0‬‬

‫ﺃﻱ ﺃﻥ ﻤﻘﺩﺍﺭ ﺘﺄﻜﺩﻨﺎ ) ﺃﻭ ﺇﺤﺘﻤﺎل( ﻤﻥ ﺃﻥ ﺍﻟﻘﻴﻤﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺍﻟﻤﺭﺍﺩ ﺍﻟﺘﻨﺒﺅ ﻋﻨﻬﺎ ﺘﺴﺎﻭﻱ ﺍﻟﻘﻴﻤﺔ‬
‫ﺍﻟﻤﻌﻁﺎﺓ ﻤﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺘﺴﺎﻭﻱ ﺍﻟﺼﻔﺭ ﺃﻱ ﺍﻨﻨﺎ ﻏﻴﺭ ﻤﺘﺄﻜﺩﻴﻥ ﺇﻁﻼﻗﺎ ﻭﺒﺎﻟﺘﺎﻟﻲ ﻻﻓﺎﺌﺩﺓ ﻤﻥ ﺍﻟﺘﻨﺒﺅ‪.‬‬
‫ﻟﻠﺘﻐﻠﺏ ﻋﻠﻰ ﺫﻟﻙ ﻭﺃﻹﺴﺘﻔﺎﺩﺓ ﻤﻥ ﺍﻟﺘﻨﺒﺅﺍﺕ ﻨﺴﺘﺨﺩﻡ ﻤﺎﻴﺴﻤﻰ ﺒﺘﻨﺒﺅ ﺍﻟﻔﺘﺭﺓ ‪Interval Forecast‬‬
‫ﻭﻫﻲ ﻋﺒﺎﺭﺓ ﻋﻥ ﻓﺘﺭﺓ ﻤﺜل ]‪ [a, b‬ﻋﻠﻰ ﺨﻁ ﺍﻷﻋﺩﺍﺩ ﺍﻟﺤﻘﻴﻘﻴﺔ ﺒﺤﻴﺙ ﻴﻜﻭﻥ‬
‫) ‪P ( a ≤ Z n +m ≤ b ) = (1 − α‬‬

‫ﻭﺒﻬﺫﺍ ﻨﺴﺘﻁﻴﻊ ﺃﻥ ﻨﺤﺩﺩ ﺩﺭﺠﺔ ﺘﺄﻜﺩﻨﺎ ﻤﻥ ﺃﻥ ﺍﻟﻘﻴﻤﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺍﻟﻤﺭﺍﺩ ﺍﻟﺘﻨﺒﺅ ﻋﻨﻬﺎ ﺘﻘﻊ ﺒﻴﻥ ﺍﻟﻘﻴﻡ ‪a‬‬
‫ﻭ ‪ b‬ﺒﺩﺭﺠﺔ ﺘﺄﻜﺩ ﺃﻭ ﺇﺤﺘﻤﺎل ‪ ) 1 − α‬ﺃﻭ ‪ ( 100 × (1 − α ) %‬ﻓﻤﺜﻼ ﻟﻭ ﻜﺎﻨﺕ ‪ α = 0.05‬ﻓﺈﻨﻨﺎ‬
‫ﻨﻜﻭﻥ ﻤﺘﺄﻜﺩﻴﻥ ﻭﺒﺈﺤﺘﻤﺎل ‪ ٩٥٪‬ﺍﻥ ﺍﻟﻘﻴﻤﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺘﻘﻊ ﺒﻴﻥ ﺍﻟﻘﻴﻡ ‪ a‬ﻭ ‪. b‬‬

‫‪٦٣‬‬

‫ﺘﻌﺭﻴﻑ ‪ :٢١‬ﻋﻠﻰ ﺇﻓﺘﺭﺍﺽ ﺃﻥ ) ‪ at N ( 0,σ 2‬ﻓﺈﻥ ﺤﺩﻭﺩ ‪ 100 × (1 − α ) %‬ﻓﺘﺭﺓ ﺘﻨﺒﺅ ﻟﻠﻘﻴﻤﺔ‬
‫ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn+l , l ≥ 1‬ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬

‫{‬

‫}‬

‫‪12‬‬

‫⎦⎤) ‪zn ( l ) ± uα 2 V ⎡⎣ en ( l‬‬
‫‪α‬‬
‫⎟⎞ ‪ 100 ⎛⎜ 1 −‬ﻟﻠﺘﻭﺯﻴﻊ )‪. N ( 0,1‬‬
‫⎠‪2‬‬
‫⎝‬

‫ﺤﻴﺙ ‪ uα 2‬ﺍﻟﻤﺌﻴﻥ‬

‫ﻓﻤﺜﻼ ﻋﻨﺩﻤﺎ ‪ α = 0.05‬ﻓﺈﻥ ‪. u0.025 = 1.96‬‬
‫ﻤﻼﺤﻅﺔ‪ :‬ﻓﻲ ﺍﻟﺘﻌﺭﻴﻑ ﺇﻓﺘﺭﻀﻨﺎ ﺃﻥ ﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﻀﺠﺔ ﺍﻟﺒﻴﻀﺎﺀ ) ‪ at N ( 0, σ 2‬ﻭﻫﺫﺍ ﻤﻤﻜﻥ‬
‫ﺇﻋﺘﻤﺎﺩﺍ ﻋﻠﻰ ﻨﻅﺭﻴﺔ ﻨﻬﺎﻴﺔ ﻤﺭﻜﺯﻴﺔ‪.‬‬
‫ﻤﺜﺎل‪ :‬ﻤﺘﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﻤﺸﺎﻫﺩﻩ ﻭﺠﺩ ﺍﻨﻬﺎ ﺘﺘﺒﻊ ﺍﻟﻨﻤﻭﺫﺝ‬
‫) ‪N ( 0,0.024‬‬

‫‪zt − 0.97 = 0.85 ( zt −1 − 0.97 ) + at , at‬‬

‫ﺇﺫﺍ ﻜﺎﻨﺕ ﺍﻟﻤﺸﺎﻫﺩﺓ ﺍﻷﺨﻴﺭﺓ ﻫﻲ ‪ ، z156 = 0.49‬ﺃﻭﺠﺩ ﺘﻨﺒﺅﺍﺕ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪z157 , z158 , z159‬‬

‫ﻭﺃﻭﺠﺩ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻟﻬﺎ ﻭﻤﻥ ﺜﻡ ﺃﻭﺠﺩ ﻓﺘﺭﺍﺕ ﺘﻨﺒﺅ ‪ ٩٥٪‬ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ‪.‬‬
‫ﺍﻟﺤل‪ :‬ﺴﺒﻕ ﺃﻥ ﺤﺴﺒﻨﺎ ﻓﻲ ﻤﺜﺎل ﺴﺎﺒﻕ ﺍﻟﺘﻨﺒﺅﺍﺕ ﻭ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻜﺎﻟﺘﺎﻟﻲ‪:‬‬
‫ﺍﻟﺘﻨﺒﺅﺍﺕ‬
‫‪z156 (1) = 0.56, z156 ( 2 ) = 0.62, z156 ( 3) = 0.68‬‬

‫ﻭﺍﻟﺘﺒﺎﻴﻨﺎﺕ‬
‫‪V ⎡⎣ e156 (1)⎤⎦ = 0.024, V ⎡⎣ e156 ( 2 )⎤⎦ = 0.041, V ⎡⎣ e156 ( 2 )⎤⎦ = 0.054‬‬
‫ﻓﺘﺭﺍﺕ ﺘﻨﺒﺅ ‪ ٩٥٪‬ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ z157 , z158 , z159‬ﻨﻭﺠﺩﻫﺎ ﻤﻥ ﺼﻴﻐﺔ ﺘﻌﺭﻴﻑ ‪٢١‬‬

‫}‬

‫‪12‬‬

‫‪= 0.56 ± 1.96 0.024 = 0.56 ± 0.304‬‬

‫}‬

‫‪12‬‬

‫{‬

‫⎦⎤) ‪zn ( l ) ± uα 2 V ⎡⎣ en ( l‬‬

‫{‬
‫‪{V ⎡⎣e‬‬
‫‪{V ⎡⎣e‬‬

‫⎦⎤)‪1 − z156 (1) ± u0.025 V ⎡⎣ e156 (1‬‬

‫‪= 0.62 ± 1.96 0.041 = 0.62 ± 0.397‬‬

‫}⎦⎤) ‪( 2‬‬

‫‪156‬‬

‫‪2 − z156 ( 2 ) ± u0.025‬‬

‫‪= 0.68 ± 1.96 0.054 = 0.68 ± 0.455‬‬

‫}⎦⎤)‪( 3‬‬

‫‪156‬‬

‫‪3 − z156 ( 3) ± u0.025‬‬

‫‪12‬‬

‫‪12‬‬

‫ﺃﻱ ﺃﻥ ) ‪ z157 ∈ ( 0.256,0.864‬ﺒﺈﺤﺘﻤﺎل ‪ ٠,٩٥‬ﻭ ﻭﻜﺫﻟﻙ ) ‪ z158 ∈ ( 0.223,1.017‬ﻭ ﻜﺫﻟﻙ ﺃﻴﻀﺎ‬
‫)‪. z159 ∈ ( 0.225,1.135‬‬

‫‪٦٤‬‬

٦٥

‫اﻟﻔﺼﻞ اﻟﺨﺎﻣﺲ‬
‫ﺗﺼﻤﻴﻢ وﺑﻨﺎء ﻧﻈﺎم ﺗﻨﺒﺆ إﺣﺼﺎﺋﻲ ‪Designing and Building‬‬
‫‪: Statistical Forecasting System‬‬
‫ﺳﺒﻖ أن ذآﺮﻧﺎ ان اﻟﺨﻄﻮة اﻷوﻟﻰ ﻟﺘﺼﻤﻴﻢ ﻧﻈﺎم ﺗﻨﺒﺆ هﻲ ﺑﻨﺎء ﻧﻤﻮذج ‪.‬إن ﻋﻤﻠﻴﺔ ﺑﻨﺎء ﻧﻤﻮذج‬
‫إﺣﺼﺎﺋﻲ هﻲ ﻋﻤﻠﻴﺔ ﺗﻜﺮارﻳﺔ ‪ Iterative‬ﺗﺘﻜﻮن ﻣﻦ ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ‪ ،‬ﺗﻘﺪﻳﺮ اﻟﻨﻤﻮذج )وﻧﻘﺼﺪ ﺑﻬﺎ‬
‫ﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ اﻟﻨﻤﻮذج( و إﺧﺘﺒﺎر اﻟﻨﻤﻮذج‪.‬‬
‫ﺗﻌﻴﻴﻦ أو ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج )‪: Model Identification (Specification‬‬
‫ﻓﻲ ﻣﺮﺣﻠﺔ ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ﻧﺴﺘﺨﺪم اﻟﺒﻴﺎﻧﺎت أو اﻟﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ ) اﻟﺘﺎرﻳﺦ( واي ﻣﻌﻠﻮﻣﺎت اﺧﺮى‬
‫ﻋﻦ اﻟﻜﻴﻔﻴﺔ اﻟﺘﻲ ﺗﻮﻟﺪت ﺑﻬﺎ اﻟﻤﺘﺴﻠﺴﻠﺔ وذﻟﻚ ﻹﻗﺘﺮاح ﻣﺠﻤﻮﻋﺔ ﻣﻦ اﻟﻨﻤﺎذج اﻟﻤﻨﺎﺳﺒﺔ‪ .‬وﻳﺘﻢ ﺗﻌﻴﻴﻦ أو‬
‫ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ﺣﺴﺐ اﻟﺨﻄﻮات اﻟﻌﺮﻳﻀﺔ اﻟﺘﺎﻟﻴﺔ‪:‬‬
‫اﻟﺨﻄﻮة اﻻوﻟﻰ‪ :‬ﺗﺤﻮﻳﻞ ﺗﺜﺒﻴﺖ اﻟﺘﺒﺎﻳﻦ ‪: Variance-stabilizing Transformation‬‬
‫ﺑﻌﺪ رﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﻲ ﻣﺨﻄﻂ زﻣﻨﻲ ‪ Time Plot‬وإﺟﺮاء ﺑﻌﺾ اﻹﺧﺘﺒﺎرات اﻹﺣﺼﺎﺋﻴﺔ ﻟﻤﻌﺮﻓﺔ‬
‫ﻓﻴﻤﺎ إذا آﺎن اﻟﺘﺒﺎﻳﻦ ﺛﺎﺑﺖ‪ ،‬وﻓﻲ ﺣﺎﻟﺔ ﻋﺪم ﺛﺒﺎت اﻟﺘﺒﺎﻳﻦ او إذا آﺎن اﻟﺘﺒﺎﻳﻦ ﻳﺘﻐﻴﺮ ﻣﻊ ﻣﺴﺘﻮى‬
‫اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﺈﻧﻨﺎ ﻧﻄﺒﻖ اﻟﺘﺤﻮﻳﻞ اﻟﻠﻮﻏﺎرﺗﻤﻲ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ وﻧﻔﺤﺼﻬﺎ ﻣﻦ ﺟﺪﻳﺪ ﻓﺈذا ﺗﻢ ﺗﺜﺒﻴﺖ‬
‫اﻟﺘﺒﺎﻳﻦ وإﻻ ﻧﻠﺠﺄ إﻟﻲ ﺗﻄﺒﻴﻖ أﺣﺪ اﻟﺘﺤﻮﻳﻼت اﻟﺘﻲ ذآﺮﻧﺎهﺎ ﻓﻲ ﺟﺪول ﺻﻔﺤﺔ ‪.٤١‬‬
‫اﻟﺨﻄﻮة اﻟﺜﺎﻧﻴﺔ‪ :‬إﺧﺘﻴﺎر درﺟﺔ اﻟﺘﻔﺮﻳﻖ ‪: d‬‬
‫إذا آﺎﻧﺖ اﻟﻤﺘﻠﺴﻠﺴﺔ أو ﺗﺤﻮﻳﻠﻬﺎ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ ﻓﻴﺠﺐ ﻋﻠﻴﻨﺎ ﺗﺤﺪﻳﺪ درﺟﺔ اﻟﺘﻔﺮﻳﻖ ‪ d‬اﻟﺘﻲ‬
‫ﺗﺠﻌﻞ اﻟﻤﻤﺘﻠﺴﻠﺴﺔ أو ﺗﺤﻮﻳﻠﻬﺎ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ وﻧﻘﻮم ﺑﺄﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻷول ﺛﻢ ﻧﻔﺤﺺ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪ -١‬اﻟﻤﺨﻄﻄﺎت اﻟﺰﻣﻨﻴﺔ ﻟﻠﻤﻤﺘﻠﺴﻠﺴﺔ أو ﺗﺤﻮﻳﻠﻬﺎ‪.‬‬
‫‪ -٢‬ﻣﺨﻄﻄﺎت داﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ‪ SACF‬و ‪. SPACF‬‬
‫‪ -٣‬إﺟﺮاء ﺗﻔﺮﻳﻖ ﺁﺧﺮ إذا اﺣﺘﺎج اﻷﻣﺮ وإﻋﺎدة اﻟﺨﻄﻮات ‪ ١‬و ‪ ٢‬اﻟﺴﺎﺑﻘﺘﻴﻦ‪.‬‬
‫اﻟﻤﺨﻄﻄﺎت اﻟﺰﻣﻨﻴﺔ ﻟﻠﻤﺘﺴﻠﺴﻼت ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ﺗﺒﻴﻦ ﺗﻐﻴﺮ ﻓﻲ اﻟﻤﺴﺘﻮى وداﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﻋﻴﻨﻲ‬
‫ﻣﺘﺨﺎﻣﺪة ﺑﺒﻂء آﻤﺎ ان داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﺗﻌﻄﻲ ﻗﻴﻤﺔ واﺣﺪة ﻗﺮﻳﺒﺔ ﻣﻦ اﻟﻮاﺣﺪ‬
‫اﻟﺼﺤﻴﺢ )ﺑﻐﺾ اﻟﻨﻈﺮ ﻋﻦ اﻹﺷﺎرة( وﺑﻘﻴﺔ اﻟﻘﻴﻢ ﻗﺮﻳﺒﺔ ﺟﺪا ﻣﻦ اﻟﺼﻔﺮ‪.‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬درﺟﺔ اﻟﺘﻔﺮﻳﻖ ‪ d‬ﻏﺎﻟﺒﺎ ﻣﺎ ﺗﻜﻮن ‪ ٠‬او ‪ ١‬او ‪. ٢‬‬
‫اﻟﺨﻄﻮة اﻟﺜﺎﻟﺜﺔ‪ :‬ﺗﺤﺪﻳﺪ ‪ p‬و ‪: q‬‬
‫ﺑﻌﺪ ان ﻧﺤﺼﻞ ﻋﻠﻰ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة ﻓﻲ آﻞ ﻣﻦ اﻟﺘﺒﺎﻳﻦ واﻟﻤﺘﻮﺳﻂ ﻧﻘﻮم ﺑﺘﺤﺪﻳﺪ درﺟﺔ اﻹﻧﺤﺪار‬
‫اﻟﺬاﺗﻲ ‪ p‬ودرﺟﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪ q‬وذﻟﻚ ﺑﻤﻘﺎرﻧﺔ أﻧﻤﺎط داﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ واﻟﺘﺮاﺑﻂ‬
‫اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﻣﻊ اﻷﻧﻤﺎط اﻟﻨﻈﺮﻳﺔ ﻟﺪاﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬
‫ﻣﺴﺘﺮﺷﺪﻳﻦ ﺑﺨﻮاص ﻧﻤﺎذج )‪ ARMA(p,q‬اﻟﻤﺬآﻮرة ﻓﻲ ﺻﻔﺤﺔ ‪ ٣٨‬واﻟﺠﺪول اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ هﺬﻩ‬
‫اﻟﺨﻮاص ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﺸﺎﺋﻌﺔ‪:‬‬

‫اﻟﻨﻤﻮذج‬
‫)‪ (١,d,٠‬و)‪AR(١‬‬
‫)‪ (٢,d,٠‬و)‪AR(٢‬‬
‫)‪ (p,d,٠‬و)‪AR(p‬‬
‫)‪ (٠,d,١‬و )‪MA(١‬‬
‫)‪ (٠,d,٢‬و )‪MA(٢‬‬

‫‪ACF‬‬
‫ﺗﺨﺎﻣﺪ اﺳﻲ أواﺳﻲ ﻣﺘﺮدد‬
‫ﺗﺨﺎﻣﺪ اﺳﻲ او ﺗﺨﺎﻣﺪ ﺟﻴﺒﻲ‬
‫ﺗﺨﺎﻣﺪ اﺳﻲ و‪ /‬او ﺗﺨﺎﻣﺪ ﺟﻴﺒﻲ‬
‫‪ρ k = 0, k > 1‬‬

‫‪PACF‬‬
‫‪φkk = 0, k > 1‬‬
‫‪φkk = 0, k > 2‬‬
‫‪φkk = 0, k > p‬‬

‫ﻳﺴﻴﻄﺮ ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ‬
‫ﻳﺴﻴﻄﺮ ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ او ﺟﻴﺒﻲ‬

‫‪ρ k = 0, k > 2‬‬

‫‪٦٦‬‬

‫)‪ (٠,d,q‬و )‪MA(q‬‬
‫)‪ (١,d,١‬و )‪ARMA(١،١‬‬
‫)‪ (p,d,q‬و )‪ARMA(p,q‬‬

‫ﻳﺴﻴﻄﺮ ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ و‪ /‬او ﺟﻴﺒﻲ‬
‫‪ρ k = 0, k > q‬‬
‫ﺗﺘﻨﺎﻗﺺ وﻳﺴﻴﻄﺮ ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ‬
‫ﺗﺘﻨﺎﻗﺺ وﺗﺘﺨﺎﻣﺪ اﺳﻴﺎ‬
‫ﻣﻦ اﻟﺘﺨﻠﻒ ‪١‬‬
‫ﻣﻦ اﻟﺘﺨﻠﻒ ‪١‬‬
‫ﺗﺘﻨﺎﻗﺺ ﺑﻌﺪ اﻟﺘﺨﻠﻒ‬
‫ﺗﺘﻨﺎﻗﺺ ﺑﻌﺪ اﻟﺘﺨﻠﻒ ‪ q - p‬وﺗﺘﺨﺎﻣﺪ‬
‫‪ p – q‬وﻳﺴﻴﻄﺮ‬
‫اﺳﻴﺎ و ‪ /‬او ﺟﻴﺒﻴﺎ ﺑﻌﺪ اﻟﺘﺨﻠﻒ ‪q – p‬‬
‫ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ‬
‫و‪/‬او ﺟﻴﺒﻲ ﺑﻌﺪ‬
‫اﻟﺘﺨﻠﻒ ‪p – q‬‬

‫اﻟﺨﻄﻮة اﻟﺮاﺑﻌﺔ‪ :‬إﺿﺎﻓﺔ ﻣﻌﻠﻢ إﻧﺠﺮاف‪:‬‬
‫إذا آﺎﻧﺖ اﻟﻤﺘﺴﻠﺴﺔ ﺗﺤﺘﺎج إﻟﻰ ﺗﻔﺮﻳﻖ ﻓﻴﺠﺐ ﻋﻠﻴﻨﺎ اﻟﺘﺄآﺪ ﻓﻴﻤﺎ إذا آﺎن ﻋﻠﻴﻨﺎ إﺿﺎﻓﺔ إﻧﺠﺮاف ﻣﻌﻠﻮم‬
‫‪ δ‬إﻟﻰ اﻟﻨﻤﻮذج وهﺬا ﻳﺘﻢ ﺑﻤﻘﺎرﻧﺔ ﻣﺘﻮﺳﻂ اﻟﻌﻴﻨﺔ ‪ w‬ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻔﺮﻗﺔ اﻟﻤﺴﺘﻘﺮة ﻣﻊ اﻟﺨﻄﺄ‬
‫اﻟﻤﻌﻴﺎري ﻟﻬﺬا اﻟﻤﺘﻮﺳﻂ‬
‫‪12‬‬

‫‪⎡c‬‬
‫⎤‬
‫⎥) ‪s.e ( w ) ≅ ⎢ 0 (1 + 2r1 + 2r2 + L + 2rK‬‬
‫‪⎣n‬‬
‫⎦‬

‫ﺣﻴﺚ ‪ c0 = γˆ0‬و ‪ r1 ,L, rK‬هﻲ اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﻌﻴﻨﻴﺔ اﻟﻤﻌﻨﻮﻳﺔ ﻟﻠﺪرﺟﺔ ‪ . K‬وﻳﻜﻮن اﻹﺧﺘﺒﺎر‬

‫‪H0 : δ = 0‬‬
‫‪H1 : δ ± 0‬‬

‫‪w‬‬
‫وﻧﺮﻓﺾ ‪ H 0‬ﻋﻨﺪ ‪ α = 0.05‬إذاآﺎﻧﺖ ‪> 1.96‬‬
‫) ‪s.e ( w‬‬

‫‪.‬‬

‫ﺗﻘﺪﻳﺮ اﻟﻨﻤﻮذج ‪: Model Estimation‬‬
‫ﺑﻌﺪ ﺗﺤﺪﻳﺪ ﺷﻜﻞ اﻟﻨﻤﻮذج ﻻﺑﺪ ﻣﻦ ﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ اﻟﻨﻤﻮذج ‪ δ‬و ‪ φ1 ,K ,φ p‬و ‪ θ1 ,K ,θ q‬و ‪ σ‬وذﻟﻚ‬
‫ﺑﺈﺳﺘﺨﺪام اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎرﻳﺨﻴﺔ اﻟﻤﺘﻮﻓﺮة ﻟﺪﻳﻨﺎ‪.‬‬
‫ﻟﻨﻔﺘﺮض ان ﻟﺪﻳﻨﺎ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪة ‪ z1 , z2 ,K, zn −1 , zn‬واﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح‬
‫‪2‬‬

‫أو‬

‫) ‪N ( 0, σ 2‬‬

‫‪φ p ( B ) wt = δ + θ q ( B ) at , at‬‬

‫‪N ( 0, σ‬‬

‫‪φ p ( B ) zt = δ + θ q ( B ) at , at‬‬

‫)‬

‫‪2‬‬

‫ﺣﻴﺚ ) ‪ φ p ( B‬و ) ‪ θ q ( B‬ﻻﻳﻮﺟﺪ ﺑﻴﻨﻬﺎ ﺟﺬور ﻣﺸﺘﺮآﺔ وﺟﺬور اﻟﻤﻌﺎدﻟﺔ ‪ φ p ( B ) = 0‬ﺗﻘﻊ ﺟﻤﻴﻌﻬﺎ‬
‫ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة ) ﺷﺮط اﻹﺳﺘﻘﺮار(‪.‬‬
‫هﻨﺎك ﻃﺮق آﺜﻴﺮة ﻟﺘﻘﺪﻳﺮ اﻟﻤﻌﺎﻟﻢ ﺳﻨﺬآﺮ ﻣﻨﻬﺎ هﻨﺎ ﻓﻘﻂ ﻃﺮﻳﻘﺘﻴﻦ ﺗﺪﺧﻞ ﺿﻤﻦ ﻧﻄﺎق هﺬا اﻟﻤﻘﺮر‬
‫وهﻤﺎ ﻃﺮﻳﻘﺔ اﻟﻌﺰوم وﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ‪.‬‬
‫أوﻻ‪ :‬ﻃﺮﻳﻘﺔ اﻟﻌﺰوم ‪: The Method of Moments‬‬
‫وﺗﻌﺘﻤﺪ ﻋﻠﻰ ﻣﺴﺎوات ﻋﺰوم اﻟﻌﻴﻨﺔ ﻣﺜﻞ ﻣﺘﻮﺳﻂ اﻟﻌﻴﻨﺔ ‪ z‬واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ ﻟﻠﻌﻴﻨﺔ ‪ rk‬ﺑﺎﻟﻌﺰوم‬
‫اﻟﻨﻈﺮﻳﺔ ﻣﺜﻞ اﻟﻤﺘﻮﺳﻂ ‪ µ‬وداﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ‪ ρ k‬وﺣﻞ اﻟﻤﻌﺎدﻻت اﻟﻨﺎﺗﺠﺔ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﻌﺎﻟﻢ اﻟﻤﺮاد‬
‫ﺗﻘﺪﻳﺮهﺎ‪.‬‬
‫ﺳﻮف ﻧﺴﺘﻌﺮض اﻟﻄﺮﻳﻘﺔ ﻟﻠﻨﻤﻮذج )‪ AR(p‬آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪n‬‬
‫‪ -١‬ﻳﻘﺪر اﻟﻤﺘﻮﺳﻂ ‪ µ‬ﺑﺎﻟﻤﻘﺪر ‪ z‬اي ‪µˆ = z = ∑i =1 zi n‬‬
‫‪ -٢‬ﻟﺘﻘﺪﻳﺮ ‪ φ1 ,K ,φ p‬ﻧﺴﺘﺨﺪم اﻟﻌﻼﻗﺔ‪:‬‬
‫‪٦٧‬‬

‫‪ρ k = φ1 ρ k −1 + φ2 ρ k − 2 + L + φ p ρ k − p , k > 1‬‬

‫واﻟﺘﻲ ﺗﻨﺘﺞ ﻣﻦ ﺿﺮب اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻟﻨﻤﻮذج )‪ AR(p‬ﺑﺎﻟﺤﺪ ‪ zt −k − µ‬وأﺧﺬ اﻟﺘﻮﻗﻊ‪ .‬ﻓﻲ‬
‫اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﺑﻮﺿﻊ ‪ k = 1, 2,K , p‬ﻧﺤﺼﻞ ﻋﻠﻰ ﻧﻈﺎم اﻟﻤﻌﺎدﻻت اﻟﻤﺴﻤﻰ ﻣﻌﺎدﻻت ﻳﻮل و‬
‫ووآﺮ ‪ Yule-Walker‬اﻟﺘﺎﻟﻲ‪:‬‬

‫‪ρ1 = φ1 + φ2 ρ1 + L + φ p ρ p −1‬‬

‫‪ρ 2 = φ1 ρ1 + φ2 + L + φ p ρ p −2‬‬
‫‪M‬‬

‫‪ρ p = φ1 ρ p −1 + φ2 ρ p −2 + L + φ p‬‬

‫و ﺑﺎﻟﺘﻌﻮﻳﺾ ﻋﻦ ‪ ρ k‬ﺑﺎﻟﻤﻘﺪر ‪ rk‬ﻧﺤﺼﻞ ﻋﻠﻰ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ φˆ1 ,K,φˆp‬آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫ﺑﻮﺿﻊ ﻣﻌﺎدﻻت ﻳﻮل و ووآﺮ ﻋﻠﻰ اﻟﺸﻜﻞ اﻟﻤﺼﻔﻮﻓﻲ‪:‬‬
‫⎞ ‪rp −1 ⎞ ⎛ φˆ1‬‬
‫⎟ ⎜‬
‫⎟ ‪rp −2 ⎟⎟ ⎜ φˆ2‬‬
‫⎟ ‪M ⎟⎜ M‬‬
‫⎟ ⎜⎟‬
‫⎟ ˆ‪1 ⎟⎠ ⎜ φ‬‬
‫⎠‪⎝ p‬‬

‫‪r2 L rp −2‬‬
‫‪r1 L rp −3‬‬
‫‪M‬‬
‫‪M‬‬
‫‪M‬‬
‫‪rp −3 L r1‬‬

‫‪r1‬‬
‫‪⎛ r1 ⎞ ⎛ 1‬‬
‫‪⎜r ⎟ ⎜ r‬‬
‫‪1‬‬
‫‪⎜ 2⎟=⎜ 1‬‬
‫‪⎜M⎟ ⎜ M‬‬
‫‪M‬‬
‫‪⎜⎜ r ⎟⎟ ⎜⎜ r‬‬
‫‪⎝ p ⎠ ⎝ p −1 rp −2‬‬

‫وﺑﺤﻞ هﺬﻩ اﻟﻤﻌﺎدﻟﺔ ﻟﻠﻤﻌﺎﻟﻢ‬
‫⎞ ‪⎛ r1‬‬
‫⎟ ‪⎜r‬‬
‫⎟‪⎜ 2‬‬
‫⎟‪⎜M‬‬
‫⎟⎟ ⎜⎜‬
‫⎠ ‪⎝ rp‬‬

‫‪−1‬‬

‫ﺗﻘﺪر ‪ σ 2‬آﺎﻟﺘﺎﻟﻲ‬

‫⎞ ‪rp −1‬‬
‫⎟⎟ ‪rp −2‬‬
‫⎟ ‪M‬‬
‫⎟‬
‫⎠⎟ ‪1‬‬

‫‪r2 L rp −2‬‬
‫‪r1 L rp −3‬‬
‫‪M‬‬
‫‪M‬‬
‫‪M‬‬
‫‪rp −3 L r1‬‬

‫)‬

‫‪r1‬‬
‫‪1‬‬
‫‪M‬‬
‫‪rp −2‬‬

‫‪⎛ φˆ1 ⎞ ⎛ 1‬‬
‫⎜ ⎟ ⎜‬
‫‪⎜ φˆ2 ⎟ ⎜ r1‬‬
‫‪⎜ M ⎟=⎜ M‬‬
‫⎜ ⎟ ⎜‬
‫‪⎜ φˆ ⎟ ⎜⎝ rp −1‬‬
‫⎠‪⎝ p‬‬

‫(‬

‫‪σˆ 2 = γˆ0 1 − φˆ1r1 − φˆ2 r2 −Lφˆp rp‬‬

‫ﺣﻴﺚ‬
‫‪1‬‬
‫‪2‬‬
‫) ‪( zt − z‬‬
‫∑‬
‫‪n t =1‬‬
‫‪n‬‬

‫هﻮ ﺗﺒﺎﻳﻦ اﻟﻌﻴﻨﺔ‪.‬‬
‫ﺗﻘﺪﻳﺮ اﻟﻌﺰوم ﻟﺒﻌﺾ اﻟﻨﻤﺎذج‪:‬‬
‫‪ -١‬ﻧﻤﻮذج )‪AR(١‬‬

‫) ‪N ( 0, σ 2‬‬

‫= ‪γˆ0‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + at , at‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ φ1‬هﻮ‬
‫‪φˆ1 = r1‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ µ‬هﻮ‬
‫‪µˆ = z‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ σ 2‬هﻮ‬

‫)‬

‫(‬

‫‪σˆ 2 = γˆ0 1 − φˆ1r1‬‬

‫ﺣﻴﺚ‬
‫‪1‬‬
‫‪2‬‬
‫) ‪( zt − z‬‬
‫∑‬
‫‪n t =1‬‬
‫‪n‬‬

‫‪٦٨‬‬

‫= ‪γˆ0‬‬

‫‪ -٢‬ﻧﻤﻮذج )‪MA(١‬‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫‪zt − µ = at − θ1at −1 , at‬‬

‫ﻹﻳﺠﺎد ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ θ1‬ﻧﺴﺘﺨﺪم اﻟﻌﻼﻗﺔ‬

‫‪−θ1‬‬
‫‪1 + θ12‬‬

‫= ‪ρ1‬‬

‫وﺑﺘﻌﻮﻳﺾ اﻟﻤﻌﺎﻟﻢ ﺑﻤﻘﺪراﺗﻬﺎ‬
‫‪−θˆ1‬‬
‫‪1 + θˆ12‬‬

‫= ‪r1‬‬

‫وﺑﺤﻞ اﻟﻤﻌﺎدﻟﺔ ﻟﻠﻤﻘﺪر ‪ θˆ1‬ﻧﺠﺪ‬
‫‪−1 ± 1 − 4r1‬‬
‫‪2r1‬‬

‫= ‪θˆ1‬‬

‫هﺬا اﻟﺤﻞ ﻳﻌﻄﻲ ﻗﻴﻤﺘﻴﻦ ﻟﻠﻤﻘﺪر ‪ θˆ1‬ﻧﺄﺧﺬ اﻟﻘﻴﻤﺔ اﻟﺘﻲ ﺗﺤﻘﻖ ‪ . θˆ1 < 1‬ﻓﻤﺜﻼ ﻟﻮ آﺎﻧﺖ ‪ r1 = −0.4‬ﻓﺈن‬
‫‪ (θˆ1 ) = −0.77‬و ‪ (θˆ1 ) = 3.27‬وﺑﺎﻟﺘﺎﻟﻲ ﻳﻜﻮن ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ θ1‬هﻮ ‪. θˆ1 = −0.77‬‬
‫‪1‬‬

‫‪2‬‬

‫‪ -٣‬ﻧﻤﻮذج )‪AR(٢‬‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + φ2 ( zt − 2 − µ ) + at , at‬‬

‫ﺑﺈﺳﺘﺨﺪام ﻣﻌﺎدﻻت ﻳﻮل ووآﺮ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ φ1‬و ‪ φ2‬هﻲ‬
‫⎞ ‪⎛ φˆ1 ⎞ ⎛ 1 r1 ⎞ −1 ⎛ r1‬‬
‫⎜=⎟ ⎜‬
‫⎠⎟ ‪⎜ φˆ ⎟ ⎝ r1 1 ⎟⎠ ⎜⎝ r2‬‬
‫⎠‪⎝ 2‬‬

‫وﻣﻨﻬﺎ ﻧﺠﺪ‬
‫‪r1 − r1r2‬‬
‫‪r −r‬‬
‫‪, φˆ2 = 2‬‬
‫‪2‬‬
‫‪1 − r1‬‬
‫‪1− r‬‬

‫‪2‬‬
‫‪1‬‬
‫‪2‬‬
‫‪1‬‬

‫= ‪φˆ1‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ µ‬هﻮ‬
‫‪µˆ = z‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ σ 2‬هﻮ‬

‫)‬

‫(‬

‫‪σˆ 2 = γˆ0 1 − φˆ1r1 − φˆ2 r2‬‬

‫ﺣﻴﺚ‬
‫‪1‬‬
‫‪2‬‬
‫) ‪( zt − z‬‬
‫∑‬
‫‪n t =1‬‬
‫‪n‬‬

‫‪ -٤‬ﻧﻤﻮذج )‪MA(٢‬‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫= ‪γˆ0‬‬

‫‪zt − µ = at − θ1at −1 − θ 2 at − 2 , at‬‬

‫ﻹﻳﺠﺎد ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ θ1‬و ‪ θ2‬ﻧﺴﺘﺨﺪم اﻟﻌﻼﻗﺎت‬

‫) ‪−θ1 (1 − θ 2‬‬
‫‪−θ 2‬‬
‫= ‪, ρ2‬‬
‫‪2‬‬
‫‪2‬‬
‫‪1 + θ1 + θ 2‬‬
‫‪1 + θ12 + θ 22‬‬

‫وﺑﺘﻌﻮﻳﺾ اﻟﻤﻘﺪرات ‪ r1‬و ‪ r2‬ﻧﺤﺼﻞ ﻋﻠﻰ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ θ1‬و ‪θ2‬‬

‫)‬

‫(‬

‫= ‪ρ1‬‬

‫‪−θˆ1 1 − θˆ2‬‬
‫‪−θˆ2‬‬
‫‪,‬‬
‫‪r‬‬
‫=‬
‫‪2‬‬
‫‪1 + θˆ12 + θˆ22‬‬
‫‪1 + θˆ12 + θˆ22‬‬

‫‪٦٩‬‬

‫= ‪r1‬‬

‫وﻧﺤﻞ ﻟﻜﻞ ﻣﻦ ‪ θˆ1‬و ‪ θˆ2‬وﻧﺄﺧﺬ اﻟﺤﻠﻮل اﻟﺘﻲ ﺗﺤﻘﻖ ‪. θ 2 − θ1 < 1, θ 2 + θ1 < 1, θ 2 < 1‬‬
‫‪ -٥‬ﻧﻤﻮذج )‪ARMA(١،١‬‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + at − θ1at −1 , at‬‬

‫ﻹﻳﺠﺎد ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ θ1‬و ‪ φ1‬ﻧﺴﺘﺨﺪم اﻟﻌﻼﻗﺎت‬

‫‪(1 − φ1θ1 )(φ1 − θ1 ) , ρ = (1 − φ1θ1 )(φ1 − θ1 ) φ‬‬
‫‪2‬‬
‫‪1‬‬
‫‪2‬‬
‫‪2‬‬
‫‪1 + θ1 − 2φ1θ1‬‬

‫‪1 + θ1 − 2φ1θ1‬‬

‫= ‪ρ1‬‬

‫وﺑﺘﻌﻮﻳﺾ اﻟﻤﻘﺪرات ‪ r1‬و ‪ r2‬ﻧﺤﺼﻞ ﻋﻠﻰ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ θ1‬و ‪φ1‬‬
‫) ˆ‪1 − φˆ θˆ )(φˆ − θ‬‬
‫(‬
‫ˆ‪φ‬‬
‫= ‪r‬‬
‫‪1‬‬

‫‪1 1‬‬

‫‪1‬‬

‫ˆ‪1 + θˆ 2 − 2φˆ θ‬‬

‫‪1‬‬

‫‪1 1‬‬

‫‪2‬‬

‫‪1‬‬

‫) ˆ‪1 − φˆ θˆ )(φˆ − θ‬‬
‫(‬
‫= ‪r‬‬
‫‪,‬‬
‫‪1‬‬

‫‪1‬‬

‫‪1 1‬‬

‫ˆ‪1 + θˆ 2 − 2φˆ θ‬‬

‫‪1 1‬‬

‫‪1‬‬

‫‪1‬‬

‫وﺑﻘﺴﻤﺔ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻟﻠﻤﻘﺪر ‪ r2‬ﻋﻠﻰ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻟﻠﻤﻘﺪر ‪ r1‬ﻧﺠﺪ‬
‫‪r2‬‬
‫‪r1‬‬

‫= ‪φˆ1‬‬

‫وهﻮ ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ . φ1‬ﻹﻳﺠﺎد ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ θ1‬ﻧﻌﻮض ﻋﻦ ‪ φˆ1‬ﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ‬
‫ﻟﻠﻤﻘﺪر ‪ r1‬ﻧﺠﺪ‬
‫⎞ ˆ ‪⎛ r2 ˆ ⎞ ⎛ r2‬‬
‫⎟ ‪⎜ 1 − r θ1 ⎟ ⎜ r − θ1‬‬
‫‪1‬‬
‫‪⎠⎝ 1‬‬
‫⎠‬
‫⎝ = ‪r1‬‬
‫‪r‬‬
‫‪1 + θˆ12 − 2 2 θˆ1‬‬
‫‪r1‬‬
‫وﻧﺤﻞ اﻟﻤﻌﺎدﻟﺔ اﻟﺘﺮﺑﻴﻌﻴﺔ اﻟﻨﺎﺗﺠﺔ ﻟﻠﻤﻘﺪر ‪ θˆ1‬و ﻧﺄﺧﺬ اﻟﻘﻴﻤﺔ اﻟﺘﻲ ﺗﺤﻘﻖ ‪. θˆ1 < 1‬‬

‫ﺗﻤﺎرﻳﻦ‪ :‬أوﺟﺪ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻤﻌﺎﻟﻢ اﻟﻨﻤﺎذج اﻟﺘﺎﻟﻴﺔ‬
‫‪ARIMA(١،١،١), ARIMA(٢،١،٠), ARIMA(٠،١،٢), ARIMA(١،٢،٠),‬‬
‫‪ARIMA(٠،٢،١), ARIMA(٠،٢،٠).‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬ﻣﻘﺪرات اﻟﻌﺰوم ﺗﺴﺘﺨﺪم آﻘﻴﻢ أوﻟﻴﺔ ﻹﻳﺠﺎد ﻣﻘﺪرات أآﺜﺮ دﻗﺔ‪.‬‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ‪Conditional Least Square‬‬
‫‪: Method‬‬
‫ﻟﻨﻤﺎذج )‪ ARMA(p,q‬واﻟﺘﻲ ﺗﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫‪φ p ( B )( zt − µ ) = θ q ( B ) at , at‬‬

‫ﺣﻴﺚ ) ‪ φ p ( B‬و ) ‪ θ q ( B‬ﻻﻳﻮﺟﺪ ﺑﻴﻨﻬﺎ ﺟﺬور ﻣﺸﺘﺮآﺔ وﺟﺬور اﻟﻤﻌﺎدﻟﺔ ‪ θ q ( B ) = 0‬ﺗﻘﻊ ﺟﻤﻴﻌﻬﺎ‬
‫ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة ) ﺷﺮط اﻹﻧﻘﻼب(‪ .‬ﺑﺈﻋﺎدة آﺘﺎﺑﺔ اﻟﻨﻤﻮذج اﻟﺴﺎﺑﻖ ﻟﻸﺧﻄﺎء ‪ at‬آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫)‪φp (B‬‬
‫)‪(z − µ‬‬
‫‪θq ( B ) t‬‬

‫= ‪at‬‬

‫اﻟﻄﺮف اﻷﻳﻤﻦ ﻳﻤﻜﻦ إﻋﺘﺒﺎرة آﺪاﻟﺔ ﻓﻲ اﻟﻤﻌﺎﻟﻢ } ‪ φ = {φ1 , φ2 ,K , φ p‬و } ‪ θ = {θ1 ,θ 2 ,K ,θ q‬و ‪µ‬‬

‫و ﻳﻜﺘﺐ‬

‫)‪) (z − µ‬‬
‫)‬
‫‪t‬‬

‫‪٧٠‬‬

‫‪p‬‬

‫‪−L − φ p B‬‬

‫‪2‬‬

‫‪p‬‬

‫‪−L −θ pB‬‬

‫‪2‬‬

‫‪(1 − φ B − φ B‬‬
‫‪(1 − θ B − θ B‬‬
‫‪2‬‬

‫‪2‬‬

‫‪1‬‬

‫‪1‬‬

‫= ) ‪at ( φ, θ, µ‬‬

‫ﺗﻌﺘﻤﺪ ﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ و ﻟﻤﺸﺎهﺪات ﻣﻌﻄﺎة } ‪ z = {z1 , z2 ,K, zn‬ﻋﻠﻰ ﺗﺼﻐﻴﺮ‬
‫اﻟﺪاﻟﺔ‬
‫‪n‬‬

‫) ‪∑ a ( φ, θ, µ z‬‬
‫‪2‬‬
‫‪t‬‬

‫= ) ‪min Sc ( φ, θ, µ‬‬
‫‪φ,θ , µ‬‬

‫‪t = p +1‬‬

‫وﺣﻞ اﻟﻤﻌﺎدﻻت اﻟﻄﺒﻴﻌﻴﺔ ‪ Normal Equations‬اﻟﻨﺎﺗﺠﺔ اﻟﺘﺎﻟﻴﺔ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﻘﺪرات‪.‬‬
‫∂‬
‫‪∂ n 2‬‬
‫) ‪Sc ( φ, θ, µ‬‬
‫=‬
‫‪∑ at ( φ, θ, µ z ) φ=φˆ = 0‬‬
‫ˆ‪φ= φ‬‬
‫‪∂φ‬‬
‫∂‬
‫‪φ‬‬
‫‪t = p +1‬‬
‫ˆ‬
‫‪θ =θ‬‬
‫ˆ‪θ =θ‬‬
‫ˆ‪µ = µ‬‬

‫ˆ‪µ = µ‬‬

‫∂‬
‫‪∂ n 2‬‬
‫) ‪Sc ( φ, θ, µ‬‬
‫=‬
‫‪∑p+1 at ( φ, θ, µ z ) φ=φˆ = 0‬‬
‫ˆ‪φ= φ‬‬
‫‪θ‬‬
‫‪∂θ‬‬
‫∂‬
‫‪t‬‬
‫=‬
‫ˆ‪θ = θ‬‬
‫ˆ‪θ =θ‬‬
‫ˆ‪µ = µ‬‬

‫ˆ‪µ = µ‬‬

‫∂‬
‫‪∂ n 2‬‬
‫) ‪Sc ( φ, θ, µ‬‬
‫=‬
‫‪∑ at ( φ, θ, µ z ) φ=φˆ = 0‬‬
‫ˆ‪φ= φ‬‬
‫‪∂µ‬‬
‫‪∂µ t = p +1‬‬
‫ˆ‪θ = θ‬‬
‫ˆ‪θ =θ‬‬
‫ˆ‪µ = µ‬‬

‫ˆ‪µ = µ‬‬

‫هﺬﻩ اﻟﻤﻘﺪرات ﺗﺴﻤﻰ ﺷﺮﻃﻴﺔ ﻷﻧﻨﺎ هﻨﺎ ﻧﺸﺘﺮط ان اﻟﻘﻴﻢ ‪ a p = a p −1 = L = a p +1− q = 0‬أي ﻣﺴﺎوﻳﺔ‬
‫ﻟﺘﻮﻗﻌﻬﺎ‪ ) .‬ﻻﺣﻆ أن اﻟﺘﺠﻤﻴﻊ ﻓﻲ اﻟﻤﻌﺎدﻻت اﻟﺴﺎﺑﻘﺔ ﻳﺒﺪأ ﻣﻦ اﻟﻘﻴﻤﺔ ‪.( t = p + 1‬‬
‫ﻳﻘﺪر اﻟﺘﺒﺎﻳﻦ ‪ σ 2‬ﻣﻦ‬

‫)‬

‫(‬

‫‪Sc φˆ , θˆ , µ‬‬

‫)‪n − ( p + q + 1‬‬

‫= ‪σˆ 2‬‬

‫ﺗﻘﺪﻳﺮات اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج‪:‬‬
‫‪ -١‬ﻧﻤﻮذج )‪AR(١‬‬

‫)‬

‫ﻟﺘﺒﺴﻴﻂ اﻹﺷﺘﻘﺎﻗﺎت ﺳﻮف ﻧﺴﺘﺒﺪل ‪ µ‬ﺑﻤﻘﺪرهﺎ ‪z‬‬

‫‪2‬‬

‫)‬

‫‪N ( 0, σ‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + at , at‬‬

‫‪N ( 0, σ‬‬

‫‪zt − z = φ1 ( zt −1 − z ) + at , at‬‬

‫‪2‬‬

‫ﻟﻤﺸﺎهﺪات ﻣﻌﻄﺎة } ‪ z = {z1 , z2 ,K, zn‬ﻧﻜﺘﺐ اﻷﺧﻄﺎء‬

‫‪at (φ1 ) = ( zt − z ) − φ1 ( zt −1 − z ) , t = 2,3,L, n‬‬

‫وﺗﺮﺑﻴﻊ اﻟﻄﺮﻓﻴﻦ واﻟﺠﻤﻊ ﻋﻠﻰ آﻞ اﻟﻤﺸﺎهﺪات‬
‫‪a (φ1 ) = ⎡⎣( zt − z ) − φ1 ( zt −1 − z )⎤⎦ , t = 2,3,L , n‬‬
‫‪2‬‬

‫‪2‬‬

‫‪2‬‬
‫‪t‬‬

‫‪n‬‬

‫‪n‬‬

‫‪t =2‬‬

‫‪t =2‬‬

‫⎦⎤) ‪Sc (φ1 ) = ∑ at2 (φ1 ) = ∑ ⎡⎣( zt − z ) − φ1 ( zt −1 − z‬‬

‫وهﺬﻩ داﻟﺔ ﻟﻠﻤﻌﻠﻢ ‪ φ1‬ﻓﻘﻂ ‪ ،‬ﻧﺸﺘﻖ اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﻌﻠﻢ ‪ φ1‬وﺗﻜﻮن اﻟﻨﺘﻴﺠﺔ ﻣﺴﺎوﻳﺔ ﻟﻠﺼﻔﺮ‬
‫ﻋﻨﺪﻣﺎ ‪ φ1 = φˆ1‬أي‬

‫‪٧١‬‬

‫‪2‬‬

‫‪n‬‬

‫‪n‬‬

‫⎦⎤) ‪Sc (φ1 ) = ∑ at2 (φ1 ) = ∑ ⎡⎣( zt − z ) − φ1 ( zt −1 − z‬‬
‫‪t =2‬‬

‫‪t =2‬‬

‫‪2‬‬
‫∂‬
‫∂‬
‫= ) ‪Sc (φ1‬‬
‫⎦⎤) ‪⎡⎣( zt − z ) − φ1 ( zt −1 − z‬‬
‫∑‬
‫‪∂φ1‬‬
‫‪∂φ1 t =2‬‬
‫‪n‬‬

‫‪n‬‬

‫⎦⎤) ‪= ∑ −2 ( zt −1 − z ) ⎡⎣ ( zt − z ) − φ1 ( zt −1 − z‬‬
‫‪t= 2‬‬

‫‪n‬‬
‫∂‬
‫‪Sc (φ1 ) ˆ = ∑ −2 ( zt −1 − z ) ⎡⎣( zt − z ) − φˆ1 ( zt −1 − z )⎦⎤ = 0‬‬
‫‪φ1 =φ1‬‬
‫‪∂φ1‬‬
‫‪t =2‬‬
‫‪n‬‬

‫‪∴ ∑ ( zt −1 − z ) ⎡⎣( zt − z ) − φˆ1 ( zt −1 − z )⎤⎦ = 0‬‬
‫‪t =2‬‬

‫‪n‬‬

‫‪n‬‬

‫‪t =2‬‬

‫‪t= 2‬‬

‫‪2‬‬
‫‪∑ ( zt −1 − z )( zt − z ) − φˆ1 ∑ ( zt −1 − z ) = 0‬‬

‫أي‬
‫) ‪− z )( zt − z‬‬

‫‪n‬‬

‫‪t −1‬‬

‫‪∑( z‬‬
‫‪t =2‬‬

‫‪2‬‬

‫)‪−z‬‬

‫‪n‬‬

‫‪t −1‬‬

‫‪∑( z‬‬

‫= ‪φˆ1‬‬

‫‪t =2‬‬

‫وهﻮ ﻣﻘﺪر اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ﻟﻠﻤﻌﻠﻢ ‪. φ1‬‬
‫ﺗﻤﺮﻳﻦ‪ :‬ﻗﺎرن ﺑﻴﻦ هﺬا اﻟﻤﻘﺪر وﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪. φ1‬‬
‫‪ -٢‬ﻧﻤﻮذج )‪MA(١‬‬

‫) ‪N ( 0,σ 2‬‬

‫‪zt − µ = at − θ1at −1 , at‬‬

‫ﻟﺘﺒﺴﻴﻂ اﻹﺷﺘﻘﺎﻗﺎت ﺳﻮف ﻧﺴﺘﺒﺪل ‪ µ‬ﺑﻤﻘﺪرهﺎ ‪ z‬وﻧﻌﻤﻞ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻌﺪﻟﺔ ﻟﻠﻤﺘﻮﺳﻂ‬
‫‪ xt = zt − z‬ﻓﻴﺼﺒﺢ اﻟﻨﻤﻮذج‬
‫) ‪N ( 0,σ 2‬‬

‫وﺑﻜﺘﺎﺑﺔ اﻟﻤﻌﺎدﻟﺔ اﻷﺧﻴﺮة ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫‪xt = at − θ1at −1 , at‬‬
‫‪at = xt − θ1at −1‬‬

‫وﻟﻤﺸﺎهﺪات ﻣﻌﻄﺎة ‪ x1 , x2 ,K , xn‬و ﺑﻮﺿﻊ ‪ a0 = 0‬ﺷﺮﻃﻴﺎ ﻧﻜﺘﺐ اﻷﺧﻄﺎء‬

‫‪a1 = x1‬‬
‫‪a2 = x2 − θ1a1‬‬
‫‪a3 = x3 − θ1a2‬‬
‫‪M‬‬
‫‪an = xn − θ1an −1‬‬

‫وﺑﺎﻟﺘﺎﻟﻲ‬
‫‪n‬‬

‫‪Sc (θ1 ) = ∑ at2‬‬
‫‪t =1‬‬

‫اﻟﺪاﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻏﻴﺮ ﺧﻄﻴﺔ ﻓﻲ اﻟﻤﻌﻠﻢ ‪ θ1‬و ﻳﻤﻜﻦ إﻳﺠﺎد ﻗﻴﻤﺔ ‪ θ1‬واﻟﺘﻲ ﺗﺼﻐﺮ ) ‪ Sc (θ1‬ﺑﻄﺮق اﻟﺒﺤﺚ‬
‫اﻟﻌﺪدﻳﺔ ﻣﺜﻞ اﻟﺒﺤﺚ اﻟﺸﺒﻜﻲ ﻓﻲ اﻟﻤﺠﺎل )‪ (-١،١‬أو إﺳﺘﺨﺪام ﻃﺮﻳﻘﺔ ﺟﺎوس‪-‬ﻧﻴﻮﺗﻦ واﻟﺘﻲ ﺗﺘﻠﺨﺺ‬
‫ﻓﻲ ﺗﻘﺮﻳﺐ ) ‪ at = at (θ1‬ﺑﺪاﻟﺔ ﺧﻄﻴﺔ ﻟﻠﻤﻌﻠﻢ ‪ θ1‬ﺣﻮل ﻗﻴﻤﺔ أوﻟﻴﺔ * ‪ θ‬ﻣﺜﻼ أي‬

‫‪٧٢‬‬

‫) * ‪dat (θ‬‬

‫اﻟﻤﺸﺘﻘﺔ‬

‫) * ‪dat (θ‬‬
‫‪d θ1‬‬

‫‪dθ1‬‬

‫) * ‪at (θ1 ) ≈ at (θ * ) + (θ1 − θ‬‬

‫ﻳﻤﻜﻦ ﺣﺴﺎﺑﻬﺎ ﺗﻜﺮارﻳﺎ وذﻟﻚ ﺑﺈﺷﺘﻘﺎق ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ ‪ at = xt − θ1at −1‬ﺑﺎﻟﻨﺴﺒﺔ‬

‫ﻟﻠﻤﻌﻠﻢ ‪ θ1‬ﻟﻨﺤﺼﻞ ﻋﻠﻰ‬

‫) ‪da0 (θ1‬‬
‫و ﺑﻘﻴﻤﺔ أوﻟﻴﺔ ‪= 0‬‬
‫‪dθ1‬‬

‫) ‪dat (θ1 ) θ1dat −1 (θ1‬‬
‫=‬
‫) ‪+ at −1 (θ1‬‬
‫‪dθ1‬‬
‫‪dθ1‬‬

‫‪ .‬اﻟﻤﻌﺎدﻟﺔ‬
‫) * ‪dat (θ‬‬
‫‪dθ1‬‬

‫)‬

‫*‬

‫‪−θ‬‬

‫‪1‬‬

‫‪) + (θ‬‬

‫*‬

‫‪at (θ1 ) ≈ at (θ‬‬

‫ﺧﻄﻴﺔ ﻓﻲ اﻟﻤﻌﻠﻢ ‪ θ1‬وﺑﺎﻟﺘﺎﻟﻲ ﺑﺎﻻﻣﻜﺎن ﺗﺼﻐﻴﺮ ﻣﺠﻤﻮع اﻟﻤﺮﺑﻌﺎت‬
‫‪n‬‬

‫‪Sc (θ1 ) = ∑ at2‬‬
‫‪t=1‬‬

‫ﺗﺤﻠﻴﻠﻴﺎ ﻟﻨﺤﺼﻞ ﻋﻠﻲ ﻣﻘﺪر ﺟﺪﻳﺪ وأﻓﻀﻞ ﻟﻠﻤﻌﻠﻢ ‪ θ1‬وﻧﻜﺮر هﺬﻩ اﻟﻌﻤﻠﻴﺔ ﺑﺈﺳﺘﺒﺪال * ‪ θ‬ﺑﺎﻟﻤﻘﺪر‬
‫اﻟﺠﺪﻳﺪ وﻧﺴﺘﻤﺮ ﺣﺘﻰ ﻳﺼﺒﺢ اﻟﻔﺮق ﺑﻴﻦ ﻣﻘﺪرﻳﻦ ﺗﺎﻟﻴﻴﻦ ﺻﻐﻴﺮ ﺟﺪا أو اﻟﻨﻘﺺ ﻓﻲ ﻣﺠﻤﻮع اﻟﻤﺮﺑﻌﺎت‬
‫ﺻﻐﻴﺮ ﺟﺪا‪ .‬ﻣﻤﻜﻦ إﺳﺘﺨﺪام ﻃﺮﻳﻘﺔ اﻟﻌﺰوم ﻹﻳﺠﺎد اﻟﻘﻴﻤﺔ أﻷوﻟﻴﺔ * ‪ θ‬ﻟﻜﻲ ﻧﺤﺼﻞ ﻋﻠﻰ ﺗﻘﺎرب‬
‫ﺳﺮﻳﻊ‪ .‬ﻃﺒﻌﺎ اﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ ﻻﺗﺘﻢ ﻳﺪوﻳﺎ ﺑﻞ ﺗﺤﺘﺎج إﻟﻰ ﺣﺎﺳﺐ ﻟﺬﻟﻚ‪.‬‬
‫ﻳﻼﺣﻆ أن ﺗﻘﺪﻳﺮ اﻟﻤﻌﺎﻟﻢ ﻟﻠﻨﻤﻮذج ﻓﻲ ﺣﺎﻟﺔ ﻧﻤﺎذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك او اﻟﻨﻤﺎذج اﻟﻤﺨﺘﻠﻄﺔ اﻟﺘﻲ‬
‫ﺗﺤﻮي ﻋﻠﻰ ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﺗﺸﻜﻞ ﺗﻌﻘﻴﺪا ﻷﻧﻬﺎ ﺗﺤﻮى ﻣﻌﺎﻟﻢ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﺑﺸﻜﻞ ﻏﻴﺮ ﺧﻄﻲ‬
‫وﻟﻬﺬا ﺗﺤﺘﺎج إﻟﻰ ﻃﺮق ﻋﺪدﻳﺔ ﻟﺤﻠﻬﺎ آﻤﺎ ﺷﺎهﺪﻧﺎ ﻓﻲ ﺣﺎﻟﺔ اﻟﻨﻤﻮذج )‪ MA(١‬وهﻮ أﺑﺴﻄﻬﺎ ﺟﻤﻴﻌﺎ‪.‬‬
‫ﺳﻮف ﻧﻜﺘﻔﻲ ﻓﻲ ﻣﻘﺮرﻧﺎ هﺬا ﻋﻠﻰ اﻟﻄﺮﻳﻘﺘﻴﻦ اﻟﺴﺎﺑﻘﺔ وﻟﻜﻦ ﻧﺬآﺮ ﺑﻌﺾ اﻟﻄﺮق اﻻﺧﺮى اﻟﻤﺴﺘﺨﺪﻣﺔ‬
‫ﻓﻲ ﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ اﻟﻨﻤﻮذج ﻣﺜﻞ‪:‬‬
‫‪ -١‬ﻃﺮﻳﻘﺔ اﻷرﺟﺤﻴﺔ اﻟﻌﻈﻤﻰ ‪Maximum Likelihood Method‬‬
‫‪ -٢‬ﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ ﻏﻴﺮ اﻟﺸﺮﻃﻴﺔ ‪Unconditional Least Squares Method‬‬
‫‪ -٣‬ﻃﺮق اﻟﺘﻘﺪﻳﺮﻏﻴﺮ اﻟﺨﻄﻴﺔ ‪Nonlinear Estimation Methods‬‬
‫ﺗﺸﺨﻴﺺ وإﺧﺘﺒﺎر اﻟﻨﻤﻮذج ‪: Model Checking and Diagnostics‬‬
‫ﺑﻌﺪ اﻟﺘﻌﺮف ﻋﻠﻰ ﻧﻤﻮذج ﻣﺒﺪﺋﻲ وﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ هﺬا اﻟﻨﻤﻮذج ﻧﺠﺮي ﺑﻌﺾ اﻟﺘﺸﺨﻴﺼﺎت ﻋﻠﻰ اﻟﺒﻮاﻗﻲ‬
‫أو أﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ )اﻧﻈﺮ ﺗﻌﺮﻳﻒ ‪ (٤‬ﻟﻨﺮى ﻣﺪى ﻣﻄﺎﺑﻘﺔ اﻟﻨﻤﻮذج ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺸﺎهﺪة ‪ ،‬وﻳﻔﺘﺮض‬
‫أن اﻟﺒﻮاﻗﻲ هﻲ ﻣﻘﺪرات ﻟﻤﺘﺴﻠﺴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪ at‬واﻟﺘﻲ ﻧﻔﺘﺮض اﻧﻬﺎ ﻣﻮزﻋﺔ ﻃﺒﻴﻌﻴﺎ ﺑﻤﺘﻮﺳﻂ‬
‫ﺻﻔﺮي وﺗﺒﺎﻳﻦ ‪ . σ 2‬اﻟﺒﻮاﻗﻲ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬
‫‪et = zt − zˆt = aˆt , t = 1, 2,..., n‬‬

‫أي ان اﻟﺒﻮاﻗﻲ هﻲ اﻟﻘﻴﻢ اﻟﻤﺸﺎهﺪة ﻧﺎﻗﺺ اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ‪.‬‬
‫ﻳﻘﻮم اﻟﺘﺸﺨﻴﺺ واﻹﺧﺘﺒﺎرات ﻋﻠﻰ ﻓﺤﺺ اﻟﺒﻮاﻗﻲ واﻟﻨﻈﺮ ﻓﻲ ﻣﺪى ﺗﺤﻘﻴﻘﻬﺎ ﻟﻔﺮﺿﻴﺎت اﻟﻨﻤﻮذج‬
‫واﻟﺘﻲ هﻲ‪:‬‬
‫‪ -١‬ﻣﺘﻮﺳﻂ ﺻﻔﺮي‬
‫‪ -٢‬اﻟﻌﺸﻮاﺋﻴﺔ‬
‫‪٧٣‬‬

‫‪ -٣‬ﻋﺪم اﻟﺘﺮاﺑﻂ‬
‫‪ -٤‬ﻣﻮزﻋﺔ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ )ﻣﺴﺘﻘﻞ وﻣﺘﻄﺎﺑﻖ ﺑﻤﺘﻮﺳﻂ ﺻﻔﺮي وﺗﺒﺎﻳﻦ ‪σ‬‬
‫أي ) ‪( at IIDN ( 0, σ 2‬‬
‫‪2‬‬

‫ﻟﻬﺬا ﻓﺈﻧﻨﺎ ﻧﺠﺮي ﺗﺸﺨﻴﺺ وهﻮ ﻣﺠﻤﻮﻋﺔ ﻣﻦ اﻹﺧﺘﺒﺎرات ﻋﻠﻲ اﻟﺒﻮاﻗﻲ ﻟﻨﺮي ﻓﻴﻤﺎ إذا آﺎﻧﺖ ﺗﺤﻘﻖ‬
‫هﺬﻩ اﻟﺸﺮوط وﻓﻲ هﺬﻩ اﻟﺤﺎﻟﺔ ﻧﻌﺘﺒﺮ اﻟﻨﻤﻮذج اﻟﻤﻄﺒﻖ ﻣﻘﺒﻮﻻ أﻣﺎ إذا ﻓﺸﻞ اﺣﺪ هﺬﻩ اﻹﺧﺘﺒﺎرات‬
‫ﻓﻴﺠﺐ ﻋﻠﻴﻨﺎ إﻋﺎدة اﻟﻨﻈﺮ وإﻗﺘﺮاح ﻧﻤﻮذج ﺁﺧﺮ‬
‫أوﻻ‪ :‬إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ‪:‬‬
‫‪H 0 : E ( at ) = 0‬‬
‫‪H 1 : E ( at ) ≠ 0‬‬

‫وهﻮ إﺧﺘﺒﺎر ﺑﺬﻳﻠﻴﻦ وﻧﺴﺘﺨﺪم ﻓﻴﺔ اﻹﺣﺼﺎﺋﺔ‬

‫‪e‬‬

‫) ‪se ( e‬‬

‫= ‪ u‬واﻟﺘﻲ ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ﻗﻴﺎﺳﻲ ﻓﻌﻨﺪ‬

‫ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬ﻧﻌﺘﺒﺮ ان ‪ E ( at ) = 0‬إذا آﺎﻧﺖ ‪ ) u < 1.96‬هﺬا ﻋﻠﻲ إﻋﺘﺒﺎر ان‬
‫ﺣﺠﻢ اﻟﻌﻴﻨﺔ اآﺒﺮ ﻣﻦ ‪ ٣٠‬وﺣﺪة وهﺬا داﺋﻤﺎ ﻣﺘﺤﻘﻖ ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﺘﻲ ﻧﺪرﺳﻬﺎ (‬
‫ﺛﺎﻧﻴﺎ‪ :‬إﺧﺘﺒﺎر اﻟﻌﺸﻮاﺋﻴﺔ‪:‬‬
‫ﻧﺨﺘﺒﺮ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ ﺑﻮاﺳﻄﺔ إﺧﺘﺒﺎر اﻟﺠﺮي ‪ Runs test‬ﺣﻮل اﻟﻤﺘﻮﺳﻂ وﺣﻮل اﻟﺼﻔﺮ‬
‫وهﻮ اﺣﺪ اﻹﺧﺘﺒﺎرات اﻟﻼﻣﻌﻠﻤﻴﺔ ) ﻳﻮﺟﺪ آﺜﻴﺮ ﻣﻦ اﻹﺧﺘﺒﺎرات ﻟﻠﻌﺸﻮاﺋﻴﺔ ﻳﺪرﺳﻬﺎ اﻟﻄﺎﻟﺐ ﻓﻲ‬
‫اﻟﻤﻘﺮر ‪ ٢٤١‬ﺑﺤﺚ وﻟﻜﻦ ﻧﻜﺘﻔﻲ هﻨﺎ ﺑﻬﺬا اﻹﺧﺘﺒﺎر(‪.‬‬
‫ﺛﺎﻟﺜﺎ‪ :‬إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ أو اﻹﺳﺘﻘﻼل‪:‬‬
‫ﻳﺨﺘﺒﺮ ﺗﺮاﺑﻂ أو إﺳﺘﻘﻼل اﻟﺒﻮاﻗﻲ ﺑﻮاﺳﻄﺔ إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ‪Autocorrelation‬‬
‫‪ test‬وذﻟﻚ ﺑﺤﺴﺎب ورﺳﻢ اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﻌﻴﻨﻴﺔ ‪ SACF‬ﻟﻠﺒﻮاﻗﻲ وﻣﻘﺎرﻧﺘﻬﺎ ﻣﻊ داﻟﺔ‬
‫اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻤﺘﺴﻠﺴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‪.‬‬
‫اﻹﺧﺘﺒﺎر‬
‫‪H 0 : ρ1 = 0‬‬

‫‪H1 : ρ1 ≠ 0‬‬
‫‪r1‬‬
‫ﺣﻴﺚ اﻹﺣﺼﺎﺋﺔ‬
‫) ‪se ( r1‬‬

‫= ‪ u‬ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ﻗﻴﺎﺳﻲ ﻓﻌﻨﺪ ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬ﻧﻌﺘﺒﺮ‬

‫ان ‪ ρ1 = 0‬إذا آﺎﻧﺖ ‪. u < 1.96‬‬
‫راﺑﻌﺎ‪ :‬إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‪:‬‬
‫ﻧﺨﺘﺒﺮ ﻓﻲ ﻣﺎ إذا آﺎﻧﺖ اﻟﺒﻮاﻗﻲ ﻣﻮزﻋﺔ ﻃﺒﻴﻌﻴﺎ وذﻟﻚ ﺑﻌﺪة ﻃﺮق ﻣﺜﻞ‪:‬‬
‫‪ -١‬إﺧﺘﺒﺎر ﺣﺴﻦ اﻟﺘﻄﺎﺑﻖ ‪ Goodness of Fit Test‬وﻧﺴﺘﺨﺪم اﻹﺧﺘﺒﺎر اﻟﻼﻣﻌﻠﻤﻲ‬
‫آﻮﻟﻤﻮﺟﻮروف‪ -‬ﺳﻤﻴﺮﻧﻮف ‪. Kolmogorov-Smirnov Test‬‬
‫‪ -٢‬ﻣﺨﻄﻂ اﻹﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ ‪. Normal Probability Plot‬‬
‫‪ -٣‬ﻣﺨﻄﻂ اﻟﺮﺑﻴﻌﺎت‪-‬اﻟﺮﺑﻴﻌﺎت ‪. Q-Q Plot‬‬
‫ﺑﻌﺾ اﻟﻤﻌﺎﻳﻴﺮ اﻻﺧﺮى ﻹﺧﺘﻴﺎر ﻧﻤﻮذج اﻟﻤﻨﺎﺳﺐ‪:‬‬
‫‪ (١‬إﺣﺼﺎﺋﻴﺔ آﻴﻮ ﻟـ ﻟﺠﻨﻖ‪-‬ﺑﻮآﺲ ‪ Ljung-Box Q statistc‬وﺗﺨﺘﺼﺮ ‪ LBQ‬وﺗﺴﺘﺨﺪم‬
‫ﻹﺧﺘﺒﺎر اﻟﻔﺮﺿﻴﺔ‪:‬‬
‫‪H 0 : ρ1 = ρ 2 = L = ρ K = 0‬‬
‫وﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬

‫‪٧٤‬‬

‫)‪χ 2 ( K − m‬‬

‫‪rk2‬‬
‫‪k =1 n − k‬‬
‫‪K‬‬

‫∑ )‪Q = n (n + 2‬‬

‫ﺣﻴﺚ ‪ m‬ﻋﺪد اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة ﻓﻲ اﻟﻨﻤﻮذج‪.‬‬
‫‪ (٢‬ﻣﻌﻴﺎر اﻹﻋﻼم اﻟﺬاﺗﻲ ‪ Automatic Information Criteria‬وﺗﺨﺘﺼﺮ ‪ AIC‬وﺗﻌﻄﻰ‬
‫ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬
‫‪2‬‬
‫‪AIC ( m ) = n ln σ a + 2m‬‬
‫ﺣﻴﺚ ‪ m‬ﻋﺪد اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة ﻓﻲ اﻟﻨﻤﻮذج وﻧﺨﺘﺎر اﻟﻨﻤﻮذج اﻟﺬي ﻳﻌﻄﻲ‬
‫) ‪min AIC ( m‬‬
‫‪m‬‬

‫أﻣﺜﻠﺔ وﺣﺎﻻت دراﺳﺔ ‪: Examples and Case Studies‬‬
‫‪ -١‬اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة‬
‫)‪z(t‬‬

‫‪59.9315‬‬
‫‪61.0344‬‬
‫‪60.6643‬‬
‫‪60.5777‬‬
‫‪59.9086‬‬
‫‪62.2658‬‬
‫‪61.9013‬‬
‫‪54.5045‬‬
‫‪55.8451‬‬
‫‪59.3572‬‬
‫‪61.8956‬‬
‫‪59.7252‬‬
‫‪58.4651‬‬
‫‪61.2797‬‬
‫‪58.0143‬‬
‫‪56.6085‬‬
‫‪56.8697‬‬
‫‪57.2961‬‬
‫‪59.6266‬‬
‫‪58.1105‬‬
‫‪61.9152‬‬
‫‪59.0778‬‬
‫‪58.2759‬‬
‫‪59.5015‬‬
‫‪63.4124‬‬

‫‪60.7196‬‬
‫‪62.8556‬‬
‫‪59.3191‬‬
‫‪59.5353‬‬
‫‪59.6004‬‬
‫‪59.1455‬‬
‫‪60.9492‬‬
‫‪58.1690‬‬
‫‪56.7437‬‬
‫‪63.8189‬‬
‫‪59.3402‬‬
‫‪59.9246‬‬
‫‪57.9624‬‬
‫‪57.8803‬‬
‫‪54.2185‬‬
‫‪57.0642‬‬
‫‪57.3940‬‬
‫‪59.3236‬‬
‫‪61.1107‬‬
‫‪58.7336‬‬
‫‪62.1957‬‬
‫‪56.9972‬‬
‫‪61.8685‬‬
‫‪56.6666‬‬
‫‪60.7356‬‬

‫‪57.2318‬‬
‫‪64.6886‬‬
‫‪60.5820‬‬
‫‪58.3755‬‬
‫‪60.1325‬‬
‫‪58.0151‬‬
‫‪59.5333‬‬
‫‪55.7339‬‬
‫‪58.9585‬‬
‫‪61.1520‬‬
‫‪59.0087‬‬
‫‪60.5289‬‬
‫‪59.1567‬‬
‫‪60.3373‬‬
‫‪55.4219‬‬
‫‪62.3728‬‬
‫‪60.1458‬‬
‫‪57.5307‬‬
‫‪61.5614‬‬
‫‪60.0377‬‬
‫‪60.8256‬‬
‫‪59.0780‬‬
‫‪63.1777‬‬
‫‪56.0309‬‬
‫‪59.2298‬‬

‫‪56.1346‬‬
‫‪63.5049‬‬
‫‪62.4654‬‬
‫‪59.3054‬‬
‫‪58.4174‬‬
‫‪59.0903‬‬
‫‪58.8802‬‬
‫‪63.7261‬‬
‫‪61.8370‬‬
‫‪61.6337‬‬
‫‪58.7564‬‬
‫‪60.8942‬‬
‫‪56.5413‬‬
‫‪61.4310‬‬
‫‪59.2086‬‬
‫‪60.8605‬‬
‫‪61.4451‬‬
‫‪53.8560‬‬
‫‪59.4119‬‬
‫‪59.3488‬‬
‫‪59.3839‬‬
‫‪59.8597‬‬
‫‪58.3583‬‬
‫‪56.1494‬‬
‫‪61.7218‬‬

‫‪56.4828‬‬
‫‪63.9622‬‬
‫‪59.1721‬‬
‫‪60.9225‬‬
‫‪58.1483‬‬
‫‪59.4554‬‬
‫‪61.7122‬‬
‫‪61.6708‬‬
‫‪58.5870‬‬
‫‪60.2990‬‬
‫‪58.2273‬‬
‫‪63.6776‬‬
‫‪54.6083‬‬
‫‪62.3827‬‬
‫‪57.8763‬‬
‫‪60.3843‬‬
‫‪63.5907‬‬
‫‪58.1711‬‬
‫‪59.9346‬‬
‫‪58.0423‬‬
‫‪55.4010‬‬
‫‪59.0997‬‬
‫‪59.5097‬‬
‫‪59.2927‬‬
‫‪61.1168‬‬

‫‪58.9275‬‬
‫‪63.1547‬‬
‫‪57.9813‬‬
‫‪61.1856‬‬
‫‪61.8108‬‬
‫‪59.1609‬‬
‫‪65.1325‬‬
‫‪62.6899‬‬
‫‪57.9363‬‬
‫‪59.9443‬‬
‫‪54.7163‬‬
‫‪60.0538‬‬
‫‪54.5550‬‬
‫‪63.3933‬‬
‫‪59.9569‬‬
‫‪60.0855‬‬
‫‪60.6919‬‬
‫‪61.1852‬‬
‫‪60.6201‬‬
‫‪60.7227‬‬
‫‪58.6501‬‬
‫‪59.0970‬‬
‫‪60.3563‬‬
‫‪60.2513‬‬
‫‪61.2179‬‬

‫‪59.5257‬‬
‫‪61.4230‬‬
‫‪58.7108‬‬
‫‪61.4761‬‬
‫‪62.1789‬‬
‫‪61.7008‬‬
‫‪60.5918‬‬
‫‪59.4444‬‬
‫‪57.3334‬‬
‫‪62.1017‬‬
‫‪57.3292‬‬
‫‪63.1070‬‬
‫‪58.1895‬‬
‫‪61.9205‬‬
‫‪56.2599‬‬
‫‪62.1362‬‬
‫‪58.4221‬‬
‫‪60.8962‬‬
‫‪60.3030‬‬
‫‪60.7021‬‬
‫‪58.5790‬‬
‫‪57.1459‬‬
‫‪60.9815‬‬
‫‪60.2052‬‬
‫‪60.9013‬‬

‫‪60.1815‬‬
‫‪61.0640‬‬
‫‪58.0059‬‬
‫‪61.2223‬‬
‫‪61.9753‬‬
‫‪60.4833‬‬
‫‪63.4411‬‬
‫‪59.3478‬‬
‫‪56.7241‬‬
‫‪58.1281‬‬
‫‪61.7840‬‬
‫‪60.9021‬‬
‫‪60.7001‬‬
‫‪61.9462‬‬
‫‪61.9448‬‬
‫‪60.9805‬‬
‫‪57.5151‬‬
‫‪59.2145‬‬
‫‪58.5278‬‬
‫‪60.3550‬‬
‫‪59.4242‬‬
‫‪60.3319‬‬
‫‪61.5555‬‬
‫‪59.7755‬‬
‫‪59.4755‬‬

‫اوﻻ ﻧﺮﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﻲ ﻣﺨﻄﻂ زﻣﻨﻲ ‪ Time Plot‬ﺑﺈﺳﺘﺨﺪام اﻟﺤﺰﻣﺔ اﻹﺣﺼﺎﺋﻴﺔ ‪MINITAB‬‬
‫آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫;')‪MTB > TSPlot 'z(t‬‬
‫>‪SUBC‬‬
‫;‪Index‬‬
‫>‪SUBC‬‬
‫;‪TDisplay 11‬‬
‫>‪SUBC‬‬
‫;‪Symbol‬‬
‫>‪SUBC‬‬
‫;‪Connect‬‬
‫>‪SUBC‬‬
‫‪Title "An obseved Time Series".‬‬

‫‪٧٥‬‬

A n o b s e v e d T im e S e r ie s

z(t)

65

60

55

50

In d e x

100

150

200

‫ﺛﺎﻧﻴﺎ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ ﺑﺈﺳﺘﺨﺪام اﻷﻣﺮ‬
MTB > %ACF 'z(t)';
SUBC>
MAXLAG 20;
SUBC>
TITLE"SACF of observed Time Series".
Executing from file: H:\MTBWIN\MACROS\ACF.MAC

Autocorrelation

S A C F o f o b s e rv e d T im e S e rie s
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

5

L a g C o rr

T

LBQ

0 .5 1
0 .2 0
-0 .0 0
-0 .0 5
-0 .0 8
-0 .1 8
-0 .1 9

7 .1 9
2 .3 2
- 0 .0 1
- 0 .5 9
- 0 .9 5
- 2 .0 5
- 2 .0 9

5 2 .4 8
6 0 .7 8
6 0 .7 8
6 1 .3 4
6 2 .8 2
6 9 .9 2
7 7 .5 8

1
2
3
4
5
6
7

10

15

L a g C o rr

T

LBQ

- 0 .1 4
- 0 .1 4
- 0 .0 9
- 0 .0 7
- 0 .0 8
- 0 .0 2
0 .0 3

-1 .5 0
-1 .5 2
-0 .9 0
-0 .7 1
-0 .7 9
-0 .2 1
0 .3 2

8 1 .7 6
8 6 .1 4
8 7 .7 3
8 8 .7 3
8 9 .9 7
9 0 .0 5
9 0 .2 7

8
9
10
11
12
13
14

20

L a g C o rr

T

LBQ

0 .0 7
0 .1 3
0 .1 7
0 .2 0
0 .1 2
0 .0 6

0 .6 8
1 .3 3
1 .7 5
2 .0 6
1 .2 1
0 .6 1

9 1 .2 3
9 4 .8 6
1 0 1 .3 3
1 1 0 .6 3
1 1 3 .9 8
1 1 4 .8 6

15
16
17
18
19
20

‫ﺛﺎﻟﺜﺎ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﺑﺈﺳﺘﺨﺪام اﻷﻣﺮ‬
MTB > %PACF 'z(t)';
SUBC>
MAXLAG 20;
SUBC>
TITLE"SPACF of obseved Time Series".
Executing from file: H:\MTBWIN\MACROS\PACF.MAC

٧٦

Partial Autocorrelation

S P A C F o f o b se ve d T im e S e rie s
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

5

10

L a g P AC

T

0 .5 1
-0 .0 8
-0 .1 0
0 .0 0
-0 .0 5
-0 .1 6
-0 .0 4

7 .1 9
-1 .0 7
-1 .3 8
0 .0 4
-0 .7 3
-2 .3 4
-0 .5 0

1
2
3
4
5
6
7

15

L a g P AC

T

-0 .0 1
-0 .1 2
0 .0 1
-0 .0 4
-0 .0 9
0 .0 3
0 .0 2

-0 .1 2
-1 .6 3
0 .1 6
-0 .6 0
-1 .3 4
0 .3 9
0 .3 2

8
9
10
11
12
13
14

20

L a g P AC

T

-0 .0 2
0 .0 9
0 .0 8
0 .0 6
-0 .0 3
0 .0 4

-0 .2 3
1 .2 8
1 .2 0
0 .8 6
-0 .4 5
0 .5 2

15
16
17
18
19
20

‫ﻣﻦ أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ و اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ ﺗﺘﺒﻊ ﻧﻤﻮذج‬
‫ وﻟﻬﺬا ﻧﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻋﻠﻲ اﻟﻤﺸﺎهﺪات ﺑﺈﺳﺘﺨﺪام اﻷﻣﺮ‬AR(١)
MTB > Name c7 = 'RESI1'
MTB > ARIMA 1 0 0 'z(t)' 'RESI1';
SUBC>
Constant;
SUBC>
Forecast 5 c4 c5 c6;
SUBC> GACF;
SUBC> GPACF;
SUBC> GHistogram;
SUBC> GNormalplot.

ARIMA Model
ARIMA model for z(t)
Estimates at each iteration
Iteration
SSE
Parameters
0
839.667
0.100
53.870
1
746.819
0.250
44.876
2
695.840
0.400
35.883
3
685.086
0.502
29.769
4
685.054
0.507
29.458
5
685.054
0.507
29.443
Relative change in each estimate less than
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.5073
0.0611
Constant
29.4429
0.1309
Mean
59.7571
0.2656

0.0010

T
8.30
224.98

Number of observations: 201
Residuals:
SS = 685.020 (backforecasts excluded)
MS =
3.442 DF = 199
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
10.8(DF=11)
27.6(DF=23)
35.9(DF=35)

48
45.0(DF=47)

Forecasts from period 201
Period
202
203

Forecast
59.7079
59.7322

95 Percent Limits
Lower
Upper
56.0707
63.3451
55.6537
63.8106

٧٧

Actual

204
205
206

59.7445
59.7507
59.7539

55.5600
55.5394
55.5357

zt = 59.76 + 0.51( zt −1 − 59.76) + at , at

63.9290
63.9620
63.9721

:‫وﻧﺴﺘﻨﺘﺞ اﻟﺘﺎﻟﻲ‬
‫ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬-١

WN ( 0,3.44 )

:‫ ﻟﻬﺎ هﻲ‬t ‫ اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة وإﻧﺤﺮاﻓﻬﺎ اﻟﻤﻌﻴﺎري و ﻗﻴﻤﺔ‬-٢

( )

φˆ1 = 0.51, s.e. φˆ1 = 0.061, t = 8.3
µˆ = 59.76, s.e. ( µˆ ) = 0.66

( )

δˆ = 29.44, s.e. δˆ = 0.131, t = 224.98
σˆ 2 = 3.44, with d . f . = 199

:‫راﺑﻌﺎ ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬
‫ إﺧﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬-١
MTB > ZTest 0.0 1.855 'RESI1';
SUBC>
Alternative 0;
SUBC> GHistogram;
SUBC> GDotplot;
SUBC> GBoxplot.

Z-Test
Test of mu = 0.000 vs mu not = 0.000
The assumed sigma = 1.85
Variable
RESI1

N
201

Mean
-0.002

StDev
1.851

SE Mean
0.131

Z
-0.01

P
0.99

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﻤﺘﻮﺳﻂ ﻳﺴﺎوي اﻟﺼﻔﺮ‬
‫ إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬-٢
MTB > Runs 0 'RESI1'.

Runs Test
RESI1
K =

0.0000

The observed number of runs = 94
The expected number of runs = 101.0796
107 Observations above K
94 below
The test is significant at 0.3149
Cannot reject at alpha = 0.05

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﺒﻮاﻗﻲ ﻋﺸﻮاﺋﻴﺔ‬
‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﺒﻮاﻗﻲ‬-٣

٧٨

ACF of Residuals for z(t)
(with 95% confidence limits for the autocorrelations)
1.0
0.8

Autocorrelation

0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
5

10

15

20

25

30

35

40

45

50

Lag

‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ‬-٤
PACF of Residuals for z(t)
(with 95% confidence limits for the partial autocorrelations)
1.0

Partial Autocorrelation

0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
5

10

15

20

25

30

35

40

45

50

Lag

‫ﻧﻼﺣﻆ ان أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﺘﺒﻊ أﻧﻤﺎط ﻣﺘﺴﻠﺴﻠﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬
: ‫ إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬-٥
‫ ﻧﺮﺳﻢ اﻟﻤﻀﻠﻊ اﻟﺘﻜﺮاري ﻟﻠﺒﻮاﻗﻲ‬-‫ا‬
Histogram of the Residuals
(response is z(t))

Frequency

30

20

10

0
-5

0

Residual

٧٩

5

:‫ وهﺬا ﻻﻳﻜﻔﻲ ﺑﻞ ﻳﺠﺐ ان ﻧﻨﻈﺮ اﻟﻰ‬.‫ﻧﻼﺣﻆ أﻧﻪ ﻣﺘﻨﺎﻇﺮ وﻟﺔ ﺷﻜﻞ اﻟﺘﻮزﻳﻊ اﻟﻄﺒﻴﻌﻲ ﺗﻘﺮﻳﺒﺎ‬
Normal Probability Plot ‫ رﺳﻢ اﻻﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ‬-‫ب‬
Normal Probability Plot for RESI1

99

Mean:

-1.6E-03

StDev:

1.85070

95
90

Percent

80
70
60
50
40
30
20
10
5

1

-5.0

-2.5

0.0

2.5

5.0

Data

:‫واﺿﺢ ﻣﻦ اﻟﺮﺳﻢ أن اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ وﻟﻠﺘﺄآﺪ ﻧﻘﻮم‬
‫ ﻟﻄﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬K-S Test ‫ ﺑﺈﺧﺘﺒﺎر‬-‫ج‬
MTB > %NormPlot 'RESI1';
SUBC>
Kstest;
SUBC>
Title "Normal Test for Residuals".
Executing from file: H:\MTBWIN\MACROS\NormPlot.MAC

Normal Test for Residuals

.999
.99

Probability

.95
.80
.50
.20
.05
.01
.001
-5

0

5

RESI1
Average: -0.0016272
StDev: 1.85070
N: 201

H 0 : Residuals

Kolmogorov-Smirnov Normality Test
D+: 0.045 D-: 0.060 D : 0.060
Approximate P-Value: 0.074

:‫وﻧﻼﺣﻆ اﻟﺘﺎﻟﻲ‬
‫اﻹﺧﺘﺒﺎر هﻮ‬

N ( 0,3.44 )

H1 : Residuals§ N ( 0,3.44 )

‫ﺳﻤﻴﺮﻧﻮف اﻋﻄﻰ‬-‫إﺧﺘﺒﺎر آﻮﻟﻤﻮﺟﻮروف‬
+

D = 0.045, D = 0.06, D = 0.06
‫ أي اﻧﻨﺎ ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ‬α = 0.05 ‫ وهﻲ أآﺒﺮ ﻣﻦ‬٠٫٠٧٤ ‫ ﻟﻺﺧﺘﺒﺎر هﻲ‬P-Value ‫اﻟـ‬

.‫اﻟﺼﻔﺮﻳﺔ‬
:‫ﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‬
٨٠

‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ‬٥ ‫اﺳﺘﺨﺪﻣﻨﺎ اﻟﻨﻤﻮذج ﻟﻠﺘﻨﺒﺆ ﻋﻦ‬
Forecasts from period 201
Period
202
203
204
205
206

Forecast
59.7079
59.7322
59.7445
59.7507
59.7539

95 Percent Limits
Lower
Upper
56.0707
63.3451
55.6537
63.8106
55.5600
63.9290
55.5394
63.9620
55.5357
63.9721

Actual

‫وﻧﺮﺳﻤﻬﺎ ﺑﺎﻷﻣﺮ اﻟﺘﺎﻟﻲ‬
Plot C4*C8 C5*C8 C6*C8;
SUBC>
Connect;
SUBC>
Type 1;
SUBC>
Color 1;
SUBC>
Size 1;
SUBC>
Title "Forecast of 5 future value with 95% limits";
SUBC>
Overlay.

Forecast of 5 future value with 95% limits
64
63
62

C4

61
60
59
58
57
56
55
1

2

3

4

5

C8

.‫واﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ اﻟﺠﺰء اﻷﺧﻴﺮ ﻣﻦ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﻊ اﻟﺘﻨﺒﺆات وﻓﺘﺮات اﻟﺘﻨﺒﺆ‬
Forecast of 5 future value with 95% limits
64
63
62

C9

61
60
59
58
57
56
55
180

190

200

C8

‫ اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة‬-٢
٨١

‫‪494.948‬‬
‫‪503.170‬‬
‫‪496.806‬‬
‫‪504.340‬‬
‫‪496.372‬‬
‫‪503.506‬‬
‫‪496.062‬‬
‫‪501.521‬‬
‫‪507.886‬‬
‫‪506.345‬‬
‫‪507.735‬‬
‫‪497.295‬‬
‫‪497.642‬‬
‫‪510.287‬‬
‫‪496.029‬‬
‫‪503.134‬‬
‫‪501.634‬‬
‫‪499.611‬‬
‫‪500.388‬‬
‫‪505.987‬‬
‫‪493.858‬‬
‫‪494.857‬‬
‫‪497.861‬‬
‫‪493.017‬‬
‫‪498.931‬‬
‫‪497.794‬‬
‫‪513.900‬‬
‫‪504.717‬‬
‫‪502.209‬‬
‫‪497.992‬‬
‫‪501.387‬‬

‫‪496.208‬‬
‫‪498.429‬‬
‫‪496.227‬‬
‫‪491.923‬‬
‫‪499.260‬‬
‫‪500.356‬‬
‫‪494.033‬‬
‫‪497.393‬‬
‫‪495.431‬‬
‫‪496.285‬‬
‫‪507.673‬‬
‫‪503.415‬‬
‫‪499.783‬‬
‫‪496.181‬‬
‫‪497.133‬‬
‫‪494.878‬‬
‫‪497.943‬‬
‫‪499.299‬‬
‫‪498.787‬‬
‫‪499.467‬‬
‫‪502.933‬‬
‫‪495.430‬‬
‫‪511.741‬‬
‫‪501.999‬‬
‫‪496.078‬‬
‫‪497.413‬‬
‫‪495.353‬‬
‫‪499.134‬‬
‫‪503.130‬‬
‫‪504.174‬‬
‫‪501.701‬‬

‫‪507.382‬‬
‫‪502.233‬‬
‫‪505.884‬‬
‫‪496.665‬‬
‫‪500.074‬‬
‫‪492.286‬‬
‫‪508.489‬‬
‫‪504.965‬‬
‫‪500.664‬‬
‫‪507.072‬‬
‫‪485.991‬‬
‫‪500.921‬‬
‫‪503.852‬‬
‫‪498.380‬‬
‫‪498.060‬‬
‫‪500.199‬‬
‫‪504.785‬‬
‫‪505.517‬‬
‫‪505.475‬‬
‫‪497.462‬‬
‫‪504.817‬‬
‫‪504.336‬‬
‫‪490.887‬‬
‫‪503.465‬‬
‫‪506.649‬‬
‫‪505.218‬‬
‫‪498.477‬‬
‫‪496.680‬‬
‫‪491.202‬‬
‫‪495.699‬‬
‫‪492.716‬‬

‫‪498.440‬‬
‫‪496.678‬‬
‫‪493.371‬‬
‫‪506.329‬‬
‫‪498.598‬‬
‫‪516.373‬‬
‫‪499.217‬‬
‫‪495.000‬‬
‫‪492.352‬‬
‫‪495.423‬‬
‫‪505.577‬‬
‫‪501.819‬‬
‫‪501.175‬‬
‫‪504.666‬‬
‫‪509.407‬‬
‫‪504.408‬‬
‫‪501.417‬‬
‫‪492.318‬‬
‫‪496.757‬‬
‫‪498.403‬‬
‫‪491.707‬‬
‫‪505.429‬‬
‫‪514.220‬‬
‫‪502.414‬‬
‫‪491.418‬‬
‫‪496.150‬‬
‫‪498.016‬‬
‫‪501.542‬‬
‫‪507.590‬‬
‫‪497.647‬‬
‫‪504.640‬‬

‫‪488.539‬‬
‫‪506.040‬‬
‫‪501.605‬‬
‫‪497.785‬‬
‫‪502.891‬‬
‫‪491.981‬‬
‫‪493.161‬‬
‫‪505.581‬‬
‫‪501.064‬‬
‫‪497.883‬‬
‫‪500.744‬‬
‫‪493.866‬‬
‫‪495.868‬‬
‫‪494.885‬‬
‫‪494.814‬‬
‫‪495.954‬‬
‫‪493.552‬‬
‫‪497.130‬‬
‫‪497.626‬‬
‫‪506.259‬‬
‫‪504.085‬‬
‫‪490.683‬‬
‫‪487.344‬‬
‫‪494.614‬‬
‫‪507.438‬‬
‫‪512.122‬‬
‫‪505.367‬‬
‫‪504.012‬‬
‫‪497.231‬‬
‫‪505.234‬‬
‫‪496.877‬‬

‫‪511.026‬‬
‫‪489.348‬‬
‫‪498.229‬‬
‫‪502.545‬‬
‫‪503.107‬‬
‫‪496.830‬‬
‫‪507.020‬‬
‫‪495.355‬‬
‫‪504.712‬‬
‫‪504.072‬‬
‫‪495.737‬‬
‫‪503.580‬‬
‫‪501.700‬‬
‫‪507.582‬‬
‫‪501.928‬‬
‫‪503.325‬‬
‫‪500.484‬‬
‫‪502.066‬‬
‫‪501.082‬‬
‫‪493.843‬‬
‫‪498.571‬‬
‫‪501.703‬‬
‫‪498.599‬‬
‫‪500.256‬‬
‫‪497.643‬‬
‫‪490.619‬‬
‫‪500.324‬‬
‫‪499.779‬‬
‫‪497.956‬‬
‫‪502.681‬‬
‫‪502.823‬‬

‫‪496.650‬‬
‫‪501.649‬‬
‫‪502.969‬‬
‫‪506.459‬‬
‫‪494.559‬‬
‫‪500.508‬‬
‫‪498.319‬‬
‫‪498.304‬‬
‫‪501.527‬‬
‫‪494.833‬‬
‫‪504.129‬‬
‫‪507.594‬‬
‫‪498.294‬‬
‫‪496.932‬‬
‫‪504.394‬‬
‫‪499.982‬‬
‫‪498.640‬‬
‫‪502.173‬‬
‫‪504.346‬‬
‫‪504.913‬‬
‫‪500.712‬‬
‫‪504.204‬‬
‫‪506.403‬‬
‫‪502.721‬‬
‫‪503.746‬‬
‫‪506.001‬‬
‫‪501.827‬‬
‫‪496.063‬‬
‫‪505.709‬‬
‫‪493.502‬‬
‫‪497.421‬‬
‫‪497.048‬‬

‫)‪z(t‬‬
‫‪499.148‬‬
‫‪503.975‬‬
‫‪498.758‬‬
‫‪493.057‬‬
‫‪499.890‬‬
‫‪507.416‬‬
‫‪498.090‬‬
‫‪504.877‬‬
‫‪494.918‬‬
‫‪499.173‬‬
‫‪496.765‬‬
‫‪482.567‬‬
‫‪500.989‬‬
‫‪501.331‬‬
‫‪489.314‬‬
‫‪502.720‬‬
‫‪502.489‬‬
‫‪495.691‬‬
‫‪496.252‬‬
‫‪499.279‬‬
‫‪498.169‬‬
‫‪497.015‬‬
‫‪503.195‬‬
‫‪500.252‬‬
‫‪498.158‬‬
‫‪500.409‬‬
‫‪496.225‬‬
‫‪491.726‬‬
‫‪489.032‬‬
‫‪500.024‬‬
‫‪505.194‬‬
‫‪499.574‬‬

‫ﺍﻭﻻ ‪ :‬ﺴﻭﻑ ﻨﺭﺴﻡ ﻓﻘﻁ ‪ ٥٠‬ﻤﺸﺎﻫﺩﺓ ﻤﻥ ﻫﺫﻩ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ‬

‫‪510‬‬

‫)‪z(t‬‬

‫‪500‬‬

‫‪490‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ‪:‬‬

‫‪٨٢‬‬

‫‪20‬‬

‫‪10‬‬

‫‪In d e x‬‬

Autocorrelation

Autocorrelation Function for z(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

5

Lag Corr

10

T

LBQ

1 -0.53 -8.38
2 -0.05 -0.64
3 0.12 1.52
4 0.04 0.52
5 -0.13 -1.61
6 0.11 1.39
7 -0.09 -1.13

71.03
71.68
75.39
75.83
80.08
83.33
85.52

Lag Corr

T

15

LBQ

8 0.10 1.19 87.98
9 -0.11 -1.31 91.01
10 0.15 1.77 96.64
11 -0.18 -2.16 105.32
12 0.11 1.25 108.35
13 0.05 0.56 108.97
14 -0.12 -1.36 112.61

Lag Corr

20

T

LBQ

15 0.05 0.60 113.33
16 -0.02 -0.26 113.47
17 0.07 0.82 114.85
18 -0.10 -1.16 117.65
19 0.07 0.75 118.81
20 -0.06 -0.73 119.93

:‫ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ‬:‫ﺛﺎﻟﺜﺎ‬
Partial Autocorrelation

Partial Autocorrelation Function for z(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

5

10

15

Lag PAC

T

Lag PAC

T

1 -0.53
2 -0.46
3 -0.29
4 -0.07
5 -0.11
6 0.02
7 -0.09

-8.38
-7.28
-4.52
-1.08
-1.71
0.30
-1.36

8 0.04
9 -0.07
10 0.12
11 -0.07
12 -0.04
13 0.08
14 -0.04

0.65
-1.09
1.94
-1.15
-0.71
1.19
-0.59

Lag PAC

20

T

15 0.03 0.49
16 -0.13 -2.09
17 0.06 1.01
18 -0.08 -1.32
19 0.03 0.44
20 -0.13 -2.07

‫ وﺑﺘﻄﺒﻴﻖ هﺬا اﻟﻨﻤﻮذج ﻧﺠﺪ‬MA(١) ‫ﻣﻦ اﻷﻧﻤﺎط اﻟﻤﺸﺎهﺪة ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﺔ ﻗﺪ ﺗﺘﺒﻊ ﻧﻤﻮذج‬
MTB > Name c7 = 'RESI1'
MTB > ARIMA 0 0 1 'z(t)' 'RESI1';
SUBC>
Constant;
SUBC>
Forecast 5 c4 c5 c6;
SUBC> GACF;
SUBC> GPACF;
SUBC> GHistogram;
SUBC> GNormalplot.

ARIMA Model
ARIMA model for z(t)
Estimates at each iteration
Iteration
SSE
Parameters
0
6081.19
0.100 500.046
1
5265.34
0.250 500.004
2
4615.22
0.400 499.980
3
4109.70
0.550 499.967
4
3766.60
0.700 499.960
5
3727.32
0.841 499.959
6
3687.70
0.797 499.963

٨٣

7
3687.08
0.790 499.962
8
3687.07
0.791 499.962
9
3687.07
0.790 499.962
Relative change in each estimate less than
Final Estimates of Parameters
Type
Coef
StDev
MA
1
0.7905
0.0386
Constant
499.962
0.051
Mean
499.962
0.051

0.0010

T
20.50
9708.40

Number of observations: 250
Residuals:
SS = 3684.13 (backforecasts excluded)
MS =
14.86 DF = 248
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
26.7(DF=11)
35.9(DF=23)
63.1(DF=35)

48
82.8(DF=47)

Forecasts from period 250
Period
251
252
253
254
255

Forecast
502.256
499.962
499.962
499.962
499.962

95 Percent Limits
Lower
Upper
494.700
509.812
490.330
509.593
490.330
509.593
490.330
509.593
490.330
509.593

zt = 499.962 + at − 0.7905at −1 , at

WN ( 0,14.86 )

Actual

:‫وﻧﺴﺘﻨﺘﺞ اﻟﺘﺎﻟﻲ‬
‫ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬-١

:‫ ﻟﻬﺎ هﻲ‬t ‫ اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة وإﻧﺤﺮاﻓﻬﺎ اﻟﻤﻌﻴﺎري و ﻗﻴﻤﺔ‬-٢

( )

θˆ1 = 0.7905, s.e. θˆ1 = 0.0386, t = 20.50

( )

µˆ = δˆ = 499.962, s.e. δˆ = 0.051, t = 9708.40
σˆ 2 = 14.86, with d . f . = 248

:‫راﺑﻌﺎ ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬
‫ إﺧﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬-١
MTB > ZTest 0.0 3.847 'RESI1';
SUBC>
Alternative 0.

Z-Test
Test of mu = 0.000 vs mu not = 0.000
The assumed sigma = 3.85
Variable
RESI1

N
250

Mean
-0.007

StDev
3.847

SE Mean
0.243

Z
-0.03

P
0.98

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﻤﺘﻮﺳﻂ ﻳﺴﺎوي اﻟﺼﻔﺮ‬
‫ إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬-٢
MTB > Runs 0 'RESI1'.

Runs Test
RESI1

٨٤

K =

0.0000

The observed number of runs = 134
The expected number of runs = 125.9920
126 Observations above K 124 below
The test is significant at 0.3103
Cannot reject at alpha = 0.05

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﺒﻮاﻗﻲ ﻋﺸﻮاﺋﻴﺔ‬
‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﺒﻮاﻗﻲ‬-٣
A C F o f R e s id u a ls fo r z (t)
( w ith 9 5 % c o n f id e n c e l im it s f o r th e a u to c o r r e l a tio n s )
1 .0

0 .8

Autocorrelation

0 .6
0 .4

0 .2
0 .0
-0 .2
-0 .4

-0 .6
-0 .8

-1 .0
5

10

15

20

25

30

35

40

45

50

55

60

Lag

‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ‬-٤
P A C F o f R e s id u a ls f o r z (t)
( w it h 9 5 % c o n f id e n c e l im it s f o r t h e p a r t ia l a u t o c o r r e l a t io n s )
1 .0

Partial Autocorrelation

0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0
5

10

15

20

25

30

35

40

45

50

55

60

Lag

‫ﻧﻼﺣﻆ ان أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﺘﺒﻊ أﻧﻤﺎط ﻣﺘﺴﻠﺴﻠﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬
: ‫ إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬-٥
‫ ﻧﺮﺳﻢ اﻟﻤﻀﻠﻊ اﻟﺘﻜﺮاري ﻟﻠﺒﻮاﻗﻲ‬-‫ا‬

٨٥

H istogram of the R esiduals
(resp on se is z(t))

Frequency

30

20

10

0
-10

0

10

R esidual

:‫ وهﺬا ﻻﻳﻜﻔﻲ ﺑﻞ ﻳﺠﺐ ان ﻧﻨﻈﺮ اﻟﻰ‬.‫ﻧﻼﺣﻆ أﻧﻪ ﻣﺘﻨﺎﻇﺮ وﻟﺔ ﺷﻜﻞ اﻟﺘﻮزﻳﻊ اﻟﻄﺒﻴﻌﻲ ﺗﻘﺮﻳﺒﺎ‬
Normal Probability Plot ‫ رﺳﻢ اﻻﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ‬-‫ب‬
Normal Probability Plot for RESI1

99

Mean:

-6.9E-03

StDev:

3.84651

95
90

Percent

80
70
60
50
40
30
20
10
5

1

-10

-5

0

5

10

Data

:‫واﺿﺢ ﻣﻦ اﻟﺮﺳﻢ أن اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ وﻟﻠﺘﺄآﺪ ﻧﻘﻮم‬
‫ ﻟﻄﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬K-S Test ‫ ﺑﺈﺧﺘﺒﺎر‬-‫ج‬
MTB > %Qqplot 'RESI1';
SUBC>
Conf 95;
SUBC>
Ci.
Executing from file: H:\MTBWIN\MACROS\Qqplot.MAC

Distribution Function Analysis
Normal Dist. Parameter Estimates
Data
Mean:
StDev:

: RESI1
-6.9E-03
3.84651

MTB > %NormPlot 'RESI1';
SUBC>
Kstest.
Executing from file: H:\MTBWIN\MACROS\NormPlot.MAC

٨٦

Normal Probability Plot

.999
.99

Probability

.95
.80
.50
.20
.05
.01
.001
-10

0

10

RESI1
Average: -0.0069004
StDev: 3.84651
N: 250

Kolmogorov-Smirnov Normality Test
D+: 0.034 D-: 0.051 D : 0.051
Approximate P-Value: 0.105

‫ﺳﻤﻴﺮﻧﻮف اﻋﻄﻰ‬-‫إﺧﺘﺒﺎر آﻮﻟﻤﻮﺟﻮروف‬
+

D = 0.034, D = 0.051, D = 0.051
‫ أي اﻧﻨﺎ ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ ان‬α = 0.05 ‫ وهﻲ أآﺒﺮ ﻣﻦ‬٠٫١٠٥ ‫ ﻟﻺﺧﺘﺒﺎر هﻲ‬P-Value ‫اﻟـ‬

.‫اﻟﺒﻮاﻗﻲ ﻣﻮزﻋﺔ ﻃﺒﻴﻌﻴﺎ‬
:‫ﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‬
‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ‬٥ ‫اﺳﺘﺨﺪﻣﻨﺎ اﻟﻨﻤﻮذج ﻟﻠﺘﻨﺒﺆ ﻋﻦ‬
Forecasts from period 250
Period
251
252
253
254
255

95 Percent Limits
Lower
Upper
494.700
509.812
490.330
509.593
490.330
509.593
490.330
509.593
490.330
509.593

Forecast
502.256
499.962
499.962
499.962
499.962

Actual

‫ ﺗﻨﺒﺆ‬٩٥% ‫واﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ اﻟﺘﻨﺒﺆات ﻣﻊ ﻓﺘﺮات‬
Forecast of 5 future values with 95% limits

C4

510

500

490
1

2

3

4

5

C8

‫ اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة‬-٣
٨٧

‫‪224.678‬‬
‫‪227.500‬‬
‫‪221.632‬‬
‫‪234.178‬‬
‫‪233.005‬‬
‫‪232.432‬‬
‫‪228.189‬‬
‫‪231.718‬‬
‫‪222.468‬‬
‫‪231.163‬‬
‫‪226.479‬‬
‫‪234.207‬‬
‫‪230.455‬‬
‫‪233.640‬‬
‫‪233.725‬‬
‫‪233.089‬‬
‫‪228.655‬‬
‫‪220.958‬‬
‫‪231.531‬‬
‫‪231.363‬‬
‫‪228.285‬‬
‫‪229.320‬‬
‫‪235.587‬‬
‫‪228.729‬‬
‫‪232.700‬‬
‫‪240.884‬‬
‫‪236.017‬‬
‫‪229.321‬‬
‫‪228.141‬‬
‫‪228.958‬‬
‫‪226.114‬‬

‫‪226.760‬‬
‫‪224.063‬‬
‫‪225.663‬‬
‫‪238.463‬‬
‫‪232.209‬‬
‫‪230.102‬‬
‫‪231.121‬‬
‫‪229.848‬‬
‫‪223.115‬‬
‫‪228.418‬‬
‫‪224.483‬‬
‫‪232.738‬‬
‫‪228.163‬‬
‫‪232.232‬‬
‫‪236.562‬‬
‫‪233.427‬‬
‫‪225.484‬‬
‫‪220.782‬‬
‫‪232.030‬‬
‫‪230.938‬‬
‫‪229.757‬‬
‫‪230.636‬‬
‫‪230.713‬‬
‫‪230.360‬‬
‫‪234.625‬‬
‫‪241.821‬‬
‫‪230.865‬‬
‫‪226.222‬‬
‫‪225.985‬‬
‫‪229.352‬‬
‫‪228.837‬‬

‫‪226.641‬‬
‫‪222.515‬‬
‫‪229.136‬‬
‫‪239.577‬‬
‫‪227.621‬‬
‫‪229.564‬‬
‫‪231.633‬‬
‫‪227.982‬‬
‫‪227.859‬‬
‫‪228.618‬‬
‫‪228.989‬‬
‫‪229.792‬‬
‫‪225.447‬‬
‫‪234.825‬‬
‫‪235.224‬‬
‫‪233.044‬‬
‫‪224.207‬‬
‫‪222.819‬‬
‫‪232.315‬‬
‫‪233.127‬‬
‫‪229.733‬‬
‫‪227.948‬‬
‫‪227.046‬‬
‫‪230.662‬‬
‫‪234.107‬‬
‫‪238.762‬‬
‫‪230.499‬‬
‫‪225.771‬‬
‫‪225.358‬‬
‫‪229.369‬‬
‫‪232.881‬‬

‫‪226.778‬‬
‫‪221.562‬‬
‫‪232.308‬‬
‫‪232.653‬‬
‫‪222.156‬‬
‫‪229.331‬‬
‫‪231.319‬‬
‫‪225.734‬‬
‫‪230.122‬‬
‫‪228.225‬‬
‫‪235.122‬‬
‫‪228.198‬‬
‫‪224.928‬‬
‫‪234.707‬‬
‫‪231.665‬‬
‫‪229.575‬‬
‫‪224.747‬‬
‫‪226.865‬‬
‫‪232.027‬‬
‫‪233.852‬‬
‫‪232.160‬‬
‫‪224.258‬‬
‫‪228.606‬‬
‫‪234.243‬‬
‫‪231.265‬‬
‫‪233.628‬‬
‫‪226.994‬‬
‫‪225.616‬‬
‫‪225.291‬‬
‫‪230.296‬‬
‫‪236.029‬‬

‫‪229.903‬‬
‫‪222.482‬‬
‫‪236.488‬‬
‫‪223.408‬‬
‫‪218.067‬‬
‫‪229.359‬‬
‫‪234.668‬‬
‫‪225.721‬‬
‫‪230.888‬‬
‫‪229.851‬‬
‫‪235.024‬‬
‫‪226.724‬‬
‫‪227.812‬‬
‫‪233.841‬‬
‫‪228.922‬‬
‫‪227.089‬‬
‫‪228.954‬‬
‫‪232.152‬‬
‫‪232.135‬‬
‫‪232.684‬‬
‫‪235.038‬‬
‫‪223.994‬‬
‫‪228.203‬‬
‫‪239.883‬‬
‫‪229.486‬‬
‫‪230.342‬‬
‫‪223.795‬‬
‫‪226.271‬‬
‫‪223.250‬‬
‫‪228.920‬‬
‫‪235.339‬‬

‫‪230.260‬‬
‫‪223.390‬‬
‫‪236.033‬‬
‫‪217.433‬‬
‫‪217.123‬‬
‫‪235.744‬‬
‫‪235.767‬‬
‫‪224.927‬‬
‫‪228.472‬‬
‫‪228.227‬‬
‫‪236.659‬‬
‫‪225.196‬‬
‫‪231.266‬‬
‫‪233.891‬‬
‫‪230.327‬‬
‫‪227.032‬‬
‫‪232.820‬‬
‫‪232.461‬‬
‫‪228.727‬‬
‫‪231.411‬‬
‫‪233.799‬‬
‫‪225.273‬‬
‫‪231.898‬‬
‫‪243.963‬‬
‫‪225.070‬‬
‫‪227.643‬‬
‫‪223.381‬‬
‫‪230.088‬‬
‫‪225.447‬‬
‫‪227.934‬‬
‫‪235.210‬‬

‫‪227.346‬‬
‫‪225.772‬‬
‫‪234.323‬‬
‫‪213.619‬‬
‫‪221.484‬‬
‫‪236.419‬‬
‫‪233.918‬‬
‫‪226.765‬‬
‫‪228.200‬‬
‫‪225.799‬‬
‫‪236.399‬‬
‫‪222.523‬‬
‫‪232.976‬‬
‫‪232.067‬‬
‫‪230.148‬‬
‫‪227.077‬‬
‫‪233.256‬‬
‫‪231.062‬‬
‫‪224.050‬‬
‫‪232.032‬‬
‫‪233.622‬‬
‫‪223.599‬‬
‫‪229.100‬‬
‫‪242.860‬‬
‫‪225.123‬‬
‫‪230.082‬‬
‫‪228.468‬‬
‫‪235.623‬‬
‫‪226.745‬‬
‫‪229.727‬‬
‫‪232.891‬‬
‫‪224.096‬‬

‫)‪z(t‬‬
‫‪229.574‬‬
‫‪224.077‬‬
‫‪230.713‬‬
‫‪215.405‬‬
‫‪228.758‬‬
‫‪234.678‬‬
‫‪234.155‬‬
‫‪227.075‬‬
‫‪230.421‬‬
‫‪224.663‬‬
‫‪233.335‬‬
‫‪223.571‬‬
‫‪234.561‬‬
‫‪232.473‬‬
‫‪231.653‬‬
‫‪230.146‬‬
‫‪233.444‬‬
‫‪230.076‬‬
‫‪221.171‬‬
‫‪230.582‬‬
‫‪232.344‬‬
‫‪224.880‬‬
‫‪227.449‬‬
‫‪239.660‬‬
‫‪225.860‬‬
‫‪229.792‬‬
‫‪235.112‬‬
‫‪238.292‬‬
‫‪227.805‬‬
‫‪230.794‬‬
‫‪231.092‬‬
‫‪225.020‬‬

‫اوﻻ ﻧﺮﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﻲ ﻣﺨﻄﻂ زﻣﻨﻲ ‪ Time Plot‬ﺑﺈﺳﺘﺨﺪام اﻟﺤﺰﻣﺔ اﻹﺣﺼﺎﺋﻴﺔ ‪MINITAB‬‬
‫آﺎﻟﺘﺎﻟﻲ‪ ٥٠) :‬ﻣﺸﺎهﺪة ﻓﻘﻂ(‬

‫‪2 4 2‬‬

‫)‪z(t‬‬

‫‪2 3 2‬‬

‫‪2 2 2‬‬
‫‪5 0‬‬

‫‪4 0‬‬

‫‪3 0‬‬

‫ﺛﺎﻧﻴﺎ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ‬

‫‪٨٨‬‬

‫‪2 0‬‬

‫‪1 0‬‬

‫‪In d e x‬‬

Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r z ( t )
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

5

Lag
1
2
3
4
5
6
7

C o rr
0
0
0
-0
-0
-0
-0

.8
.5
.1
.1
.3
.3
.2

4
1
5
6
3
3
0

T
13
5
1
-1
-2
-2
-1

.2
.2
.3
.4
.9
.8
.7

7
0
5
6
8
7
3

10

LBQ
1
2
2
2
2
3
3

7
4
4
5
8
1
2

8
4
9
6
3
1
1

.2
.2
.7
.2
.7
.2
.9

1
9
1
1
5
3
5

Lag

1
1
1
1
1

C o rr

8 -0 .0 2 -0
9 0 .1 4 1
0 0 .2 3 1
1 0 .2 3 1
2 0 .1 5 1
3 0 .0 2 0
4 -0 .1 2 -0

T
.1
.1
.9
.8
.2
.1
.9

7
8
5
8
2
5
6

15

LBQ
3
3
3
3
3
3
3

2
2
4
5
6
6
6

2
7
1
5
1
1
5

.0
.2
.4
.1
.1
.2
.0

Lag

5
4
8
6
7
5
4

1
1
1
1
1
2

5
6
7
8
9
0

C o rr
-0
-0
-0
-0
0
0

.2
.2
.1
.1
.0
.0

1
2
8
0
0
8

20

T
-1
-1
-1
-0
0
0

.6
.7
.4
.7
.0
.6

5
9
2
6
2
1

LBQ
3
3
3
4
4
4

7
8
9
0
0
0

6
9
8
1
1
3

.3
.9
.8
.3
.3
.0

2
3
1
7
7
5

‫ﺛﺎﻟﺜﺎ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ‬

Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r z ( t )
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

5

Lag
1
2
3
4
5
6
7

PAC

10

T

0 .8 4 1 3 .2 7
-0 .6 6 -1 0 .3 9
-1 .1 3
-0 .0 7
-1 .0 0
-0 .0 6
1 .7 8
0 .1 1
2 .1 8
0 .1 4
-0 .2 2
-0 .0 1

Lag

15

PAC

T

8 0 .0 7
9 -0 .0 7
1 0 0 .0 5
1 1 -0 .0 7
1 2 0 .0 6
1 3 -0 .1 5
1 4 0 .0 1

1 .0 6
-1 .1 2
0 .7 3
-1 .1 8
0 .9 0
-2 .4 1
0 .1 8

Lag

20

PAC

T

1 5 0 .0 4
1 6 -0 .0 6
1 7 0 .0 4
1 8 -0 .0 7
1 9 0 .0 9
2 0 -0 .0 9

0 .6 7
-0 .9 3
0 .6 0
-1 .1 4
1 .4 4
-1 .3 9

‫ﻣﻦ أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ و اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ ﺗﺘﺒﻊ ﻧﻤﻮذج‬
‫ وﻟﻬﺬا ﻧﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻋﻠﻲ اﻟﻤﺸﺎهﺪات‬AR(٢)
MTB > Name c7 = 'RESI1'
MTB > ARIMA 2 0 0 'z(t)' 'RESI1';
SUBC>
Constant;
SUBC>
Forecast 10 c4 c5 c6;
SUBC> GACF;
SUBC> GPACF;
SUBC> GHistogram;
SUBC> GNormalplot.

ARIMA Model
ARIMA model for z(t)
Estimates at each iteration
Iteration
SSE
Parameters
0
4257.23
0.100
0.100
1
3528.31
0.250
0.012
2
2889.23
0.400
-0.076
3
2338.97
0.550
-0.165
4
1877.39
0.700
-0.253
5
1504.46
0.850
-0.342
6
1220.13
1.000
-0.430

٨٩

183.784
169.535
155.360
141.201
127.051
112.913
98.789

7
1024.34
1.150
8
916.97
1.300
9
894.38
1.402
10
894.31
1.408
11
894.31
1.408
Relative change in each estimate

-0.519
84.690
-0.608
70.623
-0.668
61.154
-0.672
60.670
-0.672
60.646
less than 0.0010

Final Estimates of Parameters
Type
Coef
StDev
AR
1
1.4079
0.0473
AR
2
-0.6720
0.0474
Constant
60.6458
0.1203
Mean
229.638
0.456

T
29.78
-14.19
504.11

Number of observations: 250
Residuals:
SS = 893.567 (backforecasts excluded)
MS =
3.618 DF = 247
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
17.5(DF=10)
27.2(DF=22)
49.7(DF=34)

48
67.7(DF=46)

Forecasts from period 250
Period
251
252
253
254
255
256
257
258
259
260

Forecast
224.939
226.747
228.725
230.296
231.177
231.363
231.033
230.442
229.833
229.372

95 Percent Limits
Lower
Upper
221.211
228.668
220.308
233.186
220.642
236.808
221.546
239.045
222.311
240.044
222.494
240.233
222.070
239.996
221.327
239.558
220.600
239.067
220.090
238.655

zt = 60.6458 + 1.4079 zt −1 − 0.672 zt −2 + at , at

( )
s.e. (φˆ ) = 0.0474,

Actual

:‫وﻧﺴﺘﻨﺘﺞ اﻟﺘﺎﻟﻲ‬
‫ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬-١

WN ( 0,3.618)

:‫ ﻟﻬﺎ هﻲ‬t ‫ اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة وإﻧﺤﺮاﻓﻬﺎ اﻟﻤﻌﻴﺎري و ﻗﻴﻤﺔ‬-٢

φˆ1 = 1.4079, s.e. φˆ1 = 0.0473, t = 29.78
φˆ2 = −0.672,

2

t = −14.19

µˆ = 229.638, s.e. ( µˆ ) = 0.456

( )

δˆ = 60.6458, s.e. δˆ = 0.1203, t = 504.11
σˆ 2 = 3.618, with d . f . = 247

:‫راﺑﻌﺎ ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬
‫ إﺧﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬-١
MTB > ZTest 0.0 3.618 'RESI1';
SUBC>
Alternative 0.

Z-Test
Test of mu = 0.000 vs mu not = 0.000
The assumed sigma = 3.62
Variable
RESI1

N
250

Mean
-0.005

StDev
1.894

SE Mean
0.229

٩٠

Z
-0.02

P
0.98

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﻤﺘﻮﺳﻂ ﻳﺴﺎوي اﻟﺼﻔﺮ‬
‫ إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬-٢
MTB > Runs 0 'RESI1'.

Runs Test
RESI1
K =

0.0000

The observed number of runs = 125
The expected number of runs = 125.8720
129 Observations above K 121 below
The test is significant at 0.9119
Cannot reject at alpha = 0.05

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﺒﻮاﻗﻲ ﻋﺸﻮاﺋﻴﺔ‬
‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﺒﻮاﻗﻲ‬-٣
A C F o f R e s id u a ls f o r z ( t)
( w it h 9 5 % c o n f id e n c e l im it s f o r t h e a u t o c o r r e l a t io n s )
1 .0
0 .8

Autocorrelation

0 .6
0 .4
0 .2
0 .0
- 0 .2
- 0 .4
- 0 .6
- 0 .8
- 1 .0
5

10

15

20

25

30

35

40

45

50

55

60

Lag

‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ‬-٤
P A C F o f R e s id u a ls f o r z (t)
( w it h 9 5 % c o n f id e n c e l im it s f o r t h e p a r t ia l a u t o c o r r e l a t io n s )
1 .0

Partial Autocorrelation

0 .8
0 .6
0 .4
0 .2
0 .0
- 0 .2
- 0 .4
- 0 .6
- 0 .8
- 1 .0
5

10

15

20

25

30

Lag

٩١

35

40

45

50

55

60

‫ﻧﻼﺣﻆ ان أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﺘﺒﻊ أﻧﻤﺎط ﻣﺘﺴﻠﺴﻠﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬
‫‪ -٥‬إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ ‪:‬‬
‫ا‪ -‬ﻧﺮﺳﻢ اﻟﻤﻀﻠﻊ اﻟﺘﻜﺮاري ﻟﻠﺒﻮاﻗﻲ‬
‫‪H istogram of the R esiduals‬‬
‫))‪(res p on s e is z(t‬‬
‫‪30‬‬

‫‪10‬‬

‫‪0‬‬
‫‪5‬‬

‫‪-5‬‬

‫‪0‬‬

‫‪R es idual‬‬

‫ﻧﻼﺣﻆ أﻧﻪ ﻣﺘﻨﺎﻇﺮ وﻟﺔ ﺷﻜﻞ اﻟﺘﻮزﻳﻊ اﻟﻄﺒﻴﻌﻲ ﺗﻘﺮﻳﺒﺎ‪ .‬وهﺬا ﻻﻳﻜﻔﻲ ﺑﻞ ﻳﺠﺐ ان ﻧﻨﻈﺮ اﻟﻰ‪:‬‬
‫ب‪ -‬رﺳﻢ اﻻﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ ‪Normal Probability Plot‬‬
‫‪N orm al P rob ab ility P lot for R E S I1‬‬

‫‪-4 .6 E -0 3‬‬

‫‪M ean:‬‬

‫‪1 .8 9 4 3 6‬‬

‫‪S tD e v:‬‬

‫‪99‬‬

‫‪95‬‬
‫‪90‬‬
‫‪80‬‬

‫‪20‬‬
‫‪10‬‬
‫‪5‬‬

‫‪1‬‬

‫‪4‬‬

‫‪2‬‬

‫‪0‬‬

‫‪D a ta‬‬

‫واﺿﺢ ﻣﻦ اﻟﺮﺳﻢ أن اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ وﻟﻠﺘﺄآﺪ ﻧﻘﻮم‪:‬‬
‫ج‪ -‬ﺑﺈﺧﺘﺒﺎر ‪ K-S Test‬ﻟﻄﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬

‫‪٩٢‬‬

‫‪-2‬‬

‫‪-4‬‬

‫‪-6‬‬

‫‪Percent‬‬

‫‪70‬‬
‫‪60‬‬
‫‪50‬‬
‫‪40‬‬
‫‪30‬‬

‫‪Frequency‬‬

‫‪20‬‬

Normal Probability Plot

.999
.99

Probability

.95
.80
.50
.20
.05
.01
.001
-5

0

5

RESI1
Average: -0.0046305
StDev: 1.89436
N: 250

Kolmogorov-Smirnov Normality Test
D+: 0.020 D-: 0.029 D : 0.029
Approximate P-Value > 0.15

‫ وهﻲ أآﺒﺮ ﻣﻦ‬٠٫١٥ ‫ ﻟﻺﺧﺘﺒﺎر هﻲ‬P-Value ‫ﺳﻤﻴﺮﻧﻮف اﻋﻄﻰ اﻟـ‬-‫إﺧﺘﺒﺎر آﻮﻟﻤﻮﺟﻮروف‬
.‫ أي اﻧﻨﺎ ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬α = 0.05
:‫ﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‬
‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ‬١٠ ‫اﺳﺘﺨﺪﻣﻨﺎ اﻟﻨﻤﻮذج ﻟﻠﺘﻨﺒﺆ ﻋﻦ‬
Forecasts from period 250
Period
251
252
253
254
255
256
257
258
259
260

95 Percent Limits
Lower
Upper
221.211
228.668
220.308
233.186
220.642
236.808
221.546
239.045
222.311
240.044
222.494
240.233
222.070
239.996
221.327
239.558
220.600
239.067
220.090
238.655

Forecast
224.939
226.747
228.725
230.296
231.177
231.363
231.033
230.442
229.833
229.372

Actual

‫ ﻓﺘﺮات ﺗﻨﺒﺆ‬٩٥% ‫واﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ اﻟﺘﻨﺒﺆات و‬
Forecast of 10 future values with 95% limits

C4

240

230

220
0

1

2

3

4

5

6

7

8

9

C8

٩٣

10

‫‪Forecast of 10 future values with 95% limits‬‬
‫‪245‬‬

‫‪235‬‬

‫‪C9‬‬
‫‪225‬‬

‫‪215‬‬
‫‪200‬‬

‫‪0‬‬

‫‪100‬‬

‫‪C8‬‬

‫‪Forecast of 10 future values with 95% limits‬‬

‫‪240‬‬

‫‪C9‬‬

‫‪230‬‬

‫‪220‬‬
‫‪50‬‬

‫‪60‬‬

‫‪30‬‬

‫‪40‬‬

‫‪20‬‬

‫‪10‬‬

‫‪0‬‬

‫‪C8‬‬

‫اﻟﺸﻜﻞ اﻷول ﻳﺒﻴﻦ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ ﺑﻜﺎﻣﻠﻬﺎ ﻣﻊ اﻟﺘﻨﺒﺆات واﻟﺸﻜﻞ اﻟﺜﺎﻧﻲ ﻟﻠﺨﻤﺴﻴﻦ ﻗﻴﻤﺔ اﻷﺧﻴﺮة ﻣﻊ‬
‫اﻟﺘﻨﺒﺆات ﻟﺘﻮﺿﻴﺢ ﺷﻜﻞ داﻟﺔ اﻟﺘﻨﺒﺆ‪.‬‬
‫ﺣﺎﻟﺔ دراﺳﺔ‪:‬‬
‫اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺘﺎﻟﻴﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻴﻮﻣﻲ ﻟﻌﺪد اﻟﺘﻠﻔﺰﻳﻮﻧﺎت اﻟﻤﻌﻴﺒﺔ ﻓﻲ ﺧﻂ إﻧﺘﺎج ﻣﺼﻨﻊ ﻣﺎ )إﻗﺮأ ﺳﻄﺮا‬
‫ﺑﺴﻄﺮ(‬
‫‪2.09‬‬
‫‪1.57‬‬
‫‪2.07‬‬
‫‪1.78‬‬

‫‪2.00‬‬
‫‪1.42‬‬
‫‪1.82‬‬
‫‪1.68‬‬

‫‪1.76‬‬
‫‪1.54‬‬
‫‪1.85‬‬
‫‪1.79‬‬

‫‪2.83‬‬
‫‪1.46‬‬
‫‪2.08‬‬
‫‪1.37‬‬

‫‪3.44‬‬
‫‪2.05‬‬
‫‪1.42‬‬
‫‪1.15‬‬

‫‪2.40‬‬
‫‪2.50‬‬
‫‪1.39‬‬
‫‪1.25‬‬

‫اﻟﻤﺨﻄﻂ اﻟﺰﻣﻨﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬

‫‪٩٤‬‬

‫‪1.95‬‬
‫‪2.25‬‬
‫‪1.18‬‬
‫‪1.61‬‬

‫‪2.70‬‬
‫‪1.58‬‬
‫‪1.27‬‬
‫‪1.77‬‬

‫‪1.54‬‬
‫‪1.25‬‬
‫‪1.08‬‬
‫‪2.91‬‬

‫‪Defects‬‬
‫‪1.20 1.50‬‬
‫‪1.89 1.80‬‬
‫‪1.40 1.51‬‬
‫‪2.32 1.23‬‬
‫‪1.84‬‬

3 .5

3 .0

Defects

2 .5
2 .0
1 .5
1 .0

In d e x

10

20

30

40

‫اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬
Autocorrelation

Autocorrelation Function for Defects
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

1

2

3

Lag

Corr

4

5

6

7

8

T

LBQ

Lag

Corr

T

LBQ

1 0.43 2.88
2 0.26 1.49
3 0.14 0.77
4 0.08 0.43
5 -0.09 -0.46
6 -0.07 -0.39
7 -0.21 -1.10

8.84
12.18
13.18
13.50
13.89
14.18
16.57

8
9
10
11

-0.11
-0.05
-0.01
-0.04

-0.57
-0.27
-0.04
-0.19

17.25
17.41
17.41
17.50

9

10

11

Partial Autocorrelation

Partial Autocorrelation Function for Defects
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

1

2

3

Lag PAC

4

5

T

1 0.43 2.88
2 0.09 0.63
3 -0.00 -0.01
4 0.00 0.00
5 -0.16 -1.07
6 0.00 0.02
7 -0.18 -1.19

6

Lag PAC

7

8

T

8 0.07 0.44
9 0.05 0.35
10 0.01 0.09
11 -0.03 -0.23

٩٥

9

10

11

:‫ واﻟﺬي ﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬AIC ‫ﻹﺧﺘﻴﺎر اﻟﻨﻤﻮذج اﻟﻤﻨﺎﺳﺐ ﺳﻮف ﻧﺴﺘﺨﺪم ﻣﻌﻴﺎر اﻹﻋﻼم اﻟﺬاﺗﻲ‬
AIC ( m ) = n ln σ a2 + 2m
‫ ﻋﺪد اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة ﻓﻲ اﻟﻨﻤﻮذج وﻧﺨﺘﺎر اﻟﻨﻤﻮذج اﻟﺬي ﻳﻌﻄﻲ‬m ‫ﺣﻴﺚ‬
min AIC ( m )
m

:‫ﺳﻮف ﻧﻄﺒﻖ اﻟﻨﻤﺎذج ﻋﻠﻲ اﻟﺘﻮاﻟﻲ‬
MTB > ARIMA 1 0 0 'Defects' 'RESI1';
SUBC>
Constant;

ARIMA Model
ARIMA model for Defects
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.4421
0.1365
Constant
0.99280
0.06999
Mean
1.7795
0.1254

T
3.24
14.19

Number of observations: 45
Residuals:
SS = 9.47811 (backforecasts excluded)
MS = 0.22042 DF = 43
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
4.9(DF=11)
8.9(DF=23)
30.9(DF=35)

48
* (DF= *)

MTB > ARIMA 2 0 0 'Defects' 'RESI2';
SUBC>
Constant;

ARIMA Model
ARIMA model for Defects
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.3999
0.1533
AR
2
0.0989
0.1531
Constant
0.89019
0.07047
Mean
1.7762
0.1406

T
2.61
0.65
12.63

Number of observations: 45
Residuals:
SS = 9.38567 (backforecasts excluded)
MS = 0.22347 DF = 42
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
4.0(DF=10)
8.1(DF=22)
28.8(DF=34)
MTB > ARIMA 1 0 1 'Defects' 'RESI3';
SUBC>
Constant;

ARIMA Model
ARIMA model for Defects
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.5983
0.2691
MA
1
0.1926
0.3294
Constant
0.71334
0.05693

T
2.22
0.58
12.53

٩٦

48
* (DF= *)

Mean

1.7759

0.1417

Number of observations: 45
Residuals:
SS = 9.39423 (backforecasts excluded)
MS = 0.22367 DF = 42
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
4.1(DF=10)
8.3(DF=22)
29.1(DF=34)

48
* (DF= *)

MTB > ARIMA 0 0 1 'Defects' 'RESI4';
SUBC>
Constant;

ARIMA Model
ARIMA model for Defects
Final Estimates of Parameters
Type
Coef
StDev
MA
1
-0.3409
0.1431
Constant
1.78480
0.09651
Mean
1.78480
0.09651

T
-2.38
18.49

Number of observations: 45
Residuals:
SS = 10.0362 (backforecasts excluded)
MS = 0.2334 DF = 43
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
8.0(DF=11)
13.2(DF=23)
35.7(DF=35)

48
* (DF= *)

MTB > ARIMA 0 0 2 'Defects' 'RESI5';
SUBC>
Constant;

ARIMA Model
ARIMA model for Defects
Final Estimates of Parameters
Type
Coef
StDev
MA
1
-0.3869
0.1516
MA
2
-0.1816
0.1516
Constant
1.7839
0.1118
Mean
1.7839
0.1118

T
-2.55
-1.20
15.96

Number of observations: 45
Residuals:
SS = 9.61059 (backforecasts excluded)
MS = 0.22882 DF = 42
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
4.6(DF=10)
9.2(DF=22)
31.0(DF=34)
MTB > ARIMA 2 0 1 'Defects' 'RESI6';
SUBC>
Constant;

ARIMA Model
ARIMA model for Defects
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.4134
1.5680

T
0.26

٩٧

48
* (DF= *)

AR
2
MA
1
Constant
Mean

0.0929
0.0136
0.87675
1.7761

0.7113
1.5749
0.07036
0.1425

0.13
0.01
12.46

Number of observations: 45
Residuals:
SS = 9.38561 (backforecasts excluded)
MS = 0.22892 DF = 41
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
4.0(DF= 9)
8.1(DF=21)
28.8(DF=33)

48
* (DF= *)

MTB > ARIMA 1 0 2 'Defects' 'RESI7';
SUBC>
Constant;

ARIMA Model
ARIMA model for Defects
* ERROR * Model cannot be estimated with these data
MTB > ARIMA 2 0 2 'Defects' 'RESI8';
SUBC>
Constant;

ARIMA Model
ARIMA model for Defects
Final Estimates of Parameters
Type
Coef
StDev
AR
1
1.6720
0.1165
AR
2
-0.7263
0.1251
MA
1
1.3199
0.0184
MA
2
-0.3196
0.0731
Constant 0.096224
0.003323
Mean
1.77238
0.06121

T
14.35
-5.80
71.63
-4.37
28.95

Number of observations: 45
Residuals:
SS = 8.33225 (backforecasts excluded)
MS = 0.20831 DF = 40
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
4.7(DF= 8)
8.9(DF=20)
29.7(DF=32)

48
* (DF= *)

:‫وﻧﻠﺨﺺ ذﻟﻚ ﺑﺎﻟﺠﺪول اﻟﺘﺎﻟﻲ‬
Model
__________
AR (1)
AR ( 2 )
MA (1)
MA ( 2 )
ARMA (1,1)
ARMA ( 2,1)
ARMA (1, 2 )
ARMA ( 2, 2 )

σˆ

2

________
0.22042
0.22347
0.23340
0.22882
0.22367
0.22892

0.20831

m
___
3
4
3
4
4
5

6

AIC
_________
−62.0499
−59.4315
−59.4751
−58.3669
−59.3913
−56.3472

−58.5928

٩٨

‫‪min AIC ( m ) = −62.0499‬‬
‫‪m‬‬

‫أي ان أﻓﻀﻞ ﻧﻤﻮذج هﻮ )‪. AR(١‬‬
‫ﻳﺘﺮك ﻟﻠﻄﺎﻟﺐ آﺘﻤﺮﻳﻦ ﻓﺤﺺ اﻟﺒﻮاﻗﻲ وﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‪.‬‬

‫ﺣﺎﻟﺔ دراﺳﺔ‪:‬‬
‫اﻟﻤﺘﺴﻠﺴﺔ اﻟﺘﺎﻟﻴﺔ هﻲ دﺧﻞ اﻟﻤﺒﻴﻌﺎت اﻟﺴﻨﻮﻳﺔ ﺑﻤﻼﻳﻴﻦ اﻟﺮﻳﺎﻻت ﻟﺸﺮآﺔ ﻣﺎ‬
‫‪5.43‬‬
‫‪3.88‬‬
‫‪3.57‬‬
‫‪2.75‬‬
‫‪6.06‬‬
‫‪5.80‬‬
‫‪5.16‬‬
‫‪6.64‬‬
‫‪6.43‬‬
‫‪7.86‬‬

‫‪3.80‬‬
‫‪4.30‬‬
‫‪3.45‬‬
‫‪4.80‬‬
‫‪6.12‬‬
‫‪6.08‬‬
‫‪5.71‬‬
‫‪7.49‬‬
‫‪7.53‬‬
‫‪7.50‬‬

‫‪4.14‬‬
‫‪5.42‬‬
‫‪1.22‬‬
‫‪3.08‬‬
‫‪5.65‬‬
‫‪4.78‬‬
‫‪5.61‬‬
‫‪6.09‬‬
‫‪5.62‬‬
‫‪8.27‬‬

‫‪4.60‬‬
‫‪3.91‬‬
‫‪3.98‬‬
‫‪5.43‬‬
‫‪5.52‬‬
‫‪5.67‬‬
‫‪5.63‬‬
‫‪6.64‬‬
‫‪7.59‬‬
‫‪8.75‬‬

‫‪4.77‬‬
‫‪5.07‬‬
‫‪2.65‬‬
‫‪4.40‬‬
‫‪4.79‬‬
‫‪4.89‬‬
‫‪5.70‬‬
‫‪4.72‬‬
‫‪7.27‬‬
‫‪8.50‬‬

‫‪3.45‬‬
‫‪3.78‬‬
‫‪3.28‬‬
‫‪3.84‬‬
‫‪6.11‬‬
‫‪4.99‬‬
‫‪5.75‬‬
‫‪6.08‬‬
‫‪9.01‬‬
‫‪7.23‬‬

‫‪3.99‬‬
‫‪6.16‬‬
‫‪4.05‬‬
‫‪5.14‬‬
‫‪6.46‬‬
‫‪6.12‬‬
‫‪6.36‬‬
‫‪6.57‬‬
‫‪7.49‬‬
‫‪7.53‬‬

‫‪5.51‬‬
‫‪4.05‬‬
‫‪4.08‬‬
‫‪4.00‬‬
‫‪4.31‬‬
‫‪6.23‬‬
‫‪8.02‬‬
‫‪7.56‬‬
‫‪6.69‬‬
‫‪7.67‬‬

‫‪5.74‬‬
‫‪2.54‬‬
‫‪4.61‬‬
‫‪1.58‬‬
‫‪4.99‬‬
‫‪5.05‬‬
‫‪7.07‬‬
‫‪6.87‬‬
‫‪7.22‬‬
‫‪8.22‬‬

‫ﻣﺨﻄﻂ زﻣﻨﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬

‫‪9‬‬
‫‪8‬‬
‫‪7‬‬
‫‪6‬‬

‫‪Sales‬‬

‫‪5‬‬
‫‪4‬‬
‫‪3‬‬
‫‪2‬‬
‫‪1‬‬
‫‪100‬‬

‫‪90‬‬

‫‪80‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻴﺔ‬

‫‪٩٩‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪In d e x‬‬

‫‪Sales‬‬
‫‪3.49‬‬
‫‪3.96‬‬
‫‪2.89‬‬
‫‪2.52‬‬
‫‪5.77‬‬
‫‪3.20‬‬
‫‪5.13‬‬
‫‪7.20‬‬
‫‪7.26‬‬
‫‪6.42‬‬

Autocorrelation

Autocorrelation Function for Sales
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

5

Lag Corr
1
2
3
4
5
6
7

0.71
0.60
0.65
0.64
0.59
0.59
0.51

T

LBQ

7.10 51.97
4.26 89.85
3.94 134.64
3.36 177.75
2.83 215.72
2.63 253.97
2.12 282.61

15

Lag Corr
8
9
10
11
12
13
14

0.56
0.49
0.49
0.51
0.42
0.38
0.45

T

LBQ

2.22 317.32
1.87 344.49
1.79 371.67
1.82 401.71
1.46 422.60
1.29 439.91
1.50 464.06

Lag Corr
15
16
17
18
19
20
21

0.41
0.35
0.31
0.30
0.36
0.31
0.26

T

25

LBQ

1.32 483.85
1.13 499.04
0.97 510.68
0.92 521.52
1.11 537.81
0.95 550.11
0.77 558.52

Lag Corr
22
23
24
25

0.22
0.17
0.21
0.25

T

LBQ

0.67 565.04
0.50 568.77
0.64 574.90
0.75 583.43

1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

5

Lag PAC

T

1 0.71 7.10
2 0.20 1.99
3 0.35 3.53
4 0.15 1.49
5 0.09 0.92
6 0.10 1.03
7 -0.13 -1.27

15

Lag PAC

25

T

Lag PAC

T

8 0.19 1.93
9 -0.17 -1.70
10 0.14 1.44
11 0.03 0.34
12 -0.15 -1.52
13 0.02 0.25
14 0.04 0.39

15 0.03
16 -0.10
17 -0.08
18 -0.01
19 0.14
20 -0.04
21 0.00

0.32
-0.96
-0.84
-0.10
1.43
-0.44
0.04

Lag PAC

T

22 -0.17 -1.68
23 -0.10 -1.00
24 0.15 1.55
25 0.03 0.34

.‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ ﺗﺪل ﻋﻠﻰ ﺗﺨﺎﻣﺪ ﺑﻄﻴﺊ ﻣﻤﺎ ﻗﺪ ﻳﺪل ﻋﻠﻰ ﻋﺪم إﺳﺘﻘﺮار ﻓﻲ اﻟﻤﺘﻮﺳﻂ‬
‫ وﻧﺮﺳﻤﻬﺎ‬wt = ∇zt ‫ﻟﻨﺠﺮب اﻟﺘﻔﺮﻳﻖ اﻷول ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬

3
2
1

w(t)

Partial Autocorrelation

Partial Autocorrelation Function for Sales

0
-1
-2
-3

In d e x

10

20

30

40

50

60

70

80

90

100

‫ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺴﺘﻘﺮة‬.‫ﺗﺒﺪو اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة اﻵن‬

١٠٠

Autocorrelation

Autocorrelation Function for w(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

2

Lag Corr

12

T

LBQ

1 -0.30 -3.00
2 -0.31 -2.86
3 0.11 0.90
4 0.04 0.31
5 -0.04 -0.33
6 0.14 1.15
7 -0.24 -1.98

9.26
19.32
20.49
20.63
20.80
22.80
29.02

Lag Corr

T

LBQ

8 0.19 1.53
9 -0.07 -0.53
10 -0.07 -0.52
11 0.16 1.27
12 -0.07 -0.53
13 -0.14 -1.04
14 0.15 1.10

33.07
33.58
34.08
37.12
37.67
39.85
42.37

Lag Corr

22

T

LBQ

15 0.00 0.00
16 0.02 0.13
17 -0.07 -0.52
18 -0.15 -1.09
19 0.20 1.48
20 0.01 0.05
21 -0.04 -0.25

42.37
42.40
42.99
45.65
50.73
50.74
50.90

Lag Corr

T

LBQ

22 0.05 0.35 51.22
23 -0.18 -1.28 55.44
24 0.03 0.20 55.55

Partial Autocorrelation

Partial Autocorrelation Function for w(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

2

12

22

Lag PAC

T

Lag PAC

T

Lag PAC

T

1 -0.30
2 -0.44
3 -0.22
4 -0.21
5 -0.17
6 0.06
7 -0.26

-3.00
-4.41
-2.23
-2.05
-1.72
0.59
-2.55

8 0.12
9 -0.16
10 -0.06
11 0.09
12 -0.06
13 -0.02
14 -0.07

1.17
-1.59
-0.60
0.93
-0.59
-0.17
-0.71

15 0.06
16 0.07
17 -0.03
18 -0.17
19 -0.03
20 -0.10
21 0.09

0.59
0.70
-0.32
-1.65
-0.31
-0.95
0.85

Lag PAC

T

22 0.08 0.80
23 -0.15 -1.53
24 -0.05 -0.55

:‫ واﻟﺬي ﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬AIC ‫ﻹﺧﺘﻴﺎر اﻟﻨﻤﻮذج اﻟﻤﻨﺎﺳﺐ ﺳﻮف ﻧﺴﺘﺨﺪم ﻣﻌﻴﺎر اﻹﻋﻼم اﻟﺬاﺗﻲ‬
AIC ( m ) = n ln σ a2 + 2m
‫ ﻋﺪد اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة ﻓﻲ اﻟﻨﻤﻮذج وﻧﺨﺘﺎر اﻟﻨﻤﻮذج اﻟﺬي ﻳﻌﻄﻲ‬m ‫ﺣﻴﺚ‬
min AIC ( m )
m

:‫ﺳﻮف ﻧﻄﺒﻖ اﻟﻨﻤﺎذج ﻋﻠﻲ اﻟﺘﻮاﻟﻲ‬
MTB > ARIMA 1 1 0 'Sales';
SUBC>
NoConstant.

ARIMA Model
ARIMA model for Sales
Final Estimates of Parameters
Type
Coef
StDev
AR
1
-0.3114
0.0959

T
-3.25

Differencing: 1 regular difference
Number of observations: Original series 100, after differencing 99
Residuals:
SS = 133.134 (backforecasts excluded)

١٠١

MS =

1.359

DF = 98

Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
31.9(DF=11)
51.2(DF=23)
62.8(DF=35)

48
81.0(DF=47)

MTB > ARIMA 2 1 0 'Sales';
SUBC>
NoConstant.

ARIMA Model
ARIMA model for Sales
Final Estimates of Parameters
Type
Coef
StDev
AR
1
-0.4532
0.0897
AR
2
-0.4656
0.0901

T
-5.05
-5.17

Differencing: 1 regular difference
Number of observations: Original series 100, after differencing 99
Residuals:
SS = 104.715 (backforecasts excluded)
MS =
1.080 DF = 97
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
21.8(DF=10)
40.9(DF=22)
49.4(DF=34)

48
59.9(DF=46)

MTB > ARIMA 0 1 1 'Sales';
SUBC>
NoConstant.

ARIMA Model
ARIMA model for Sales
Final Estimates of Parameters
Type
Coef
StDev
MA
1
0.7636
0.0648

T
11.78

Differencing: 1 regular difference
Number of observations: Original series 100, after differencing 99
Residuals:
SS = 101.411 (backforecasts excluded)
MS =
1.035 DF = 98
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
12.6(DF=11)
27.8(DF=23)
35.9(DF=35)

48
48.5(DF=47)

MTB > ARIMA 0 1 2 'Sales';
SUBC>
NoConstant.

ARIMA Model
ARIMA model for Sales
Final Estimates of Parameters
Type
Coef
StDev
MA
1
0.5756
0.0990
MA
2
0.2029
0.0998

T
5.81
2.03

Differencing: 1 regular difference
Number of observations: Original series 100, after differencing 99
Residuals:
SS = 99.2463 (backforecasts excluded)
MS = 1.0232 DF = 97

١٠٢

Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
14.3(DF=10)
28.3(DF=22)
36.5(DF=34)

48
47.0(DF=46)

MTB > ARIMA 1 1 1 'Sales';
SUBC>
NoConstant.

ARIMA Model
ARIMA model for Sales
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.1283
0.1334
MA
1
0.8027
0.0799

T
0.96
10.04

Differencing: 1 regular difference
Number of observations: Original series 100, after differencing 99
Residuals:
SS = 100.421 (backforecasts excluded)
MS =
1.035 DF = 97
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
13.3(DF=10)
27.9(DF=22)
36.1(DF=34)

48
48.2(DF=46)

MTB > ARIMA 2 1 1 'Sales';
SUBC>
NoConstant.

ARIMA Model
ARIMA model for Sales
* WARNING * Back forecasts not dying out rapidly
Final Estimates of Parameters
Type
Coef
StDev
AR
1
-1.1389
0.0987
AR
2
-0.1440
0.0983
MA
1
-0.9889
0.0002

T
-11.53
-1.47
-3987.49

Differencing: 1 regular difference
Number of observations: Original series 100, after differencing 99
Residuals:
SS = 134.250 (backforecasts excluded)
MS =
1.398 DF = 96
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
35.1(DF= 9)
53.5(DF=21)
66.6(DF=33)
MTB > ARIMA 1 1 2 'Sales';
SUBC>
NoConstant.

ARIMA Model
ARIMA model for Sales
Final Estimates of Parameters
Type
Coef
StDev
AR
1
-0.3476
0.4077
MA
1
0.2422
0.3771
MA
2
0.4506
0.2656

T
-0.85
0.64
1.70

Differencing: 1 regular difference

١٠٣

48
83.2(DF=45)

Number of observations: Original series 100, after differencing 99
Residuals:
SS = 97.2357 (backforecasts excluded)
MS = 1.0129 DF = 96
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
11.9(DF= 9)
25.0(DF=21)
32.1(DF=33)

48
41.8(DF=45)

MTB > ARIMA 2 1 2 'Sales';
SUBC>
NoConstant.

ARIMA Model
ARIMA model for Sales
Final Estimates of Parameters
Type
Coef
StDev
AR
1
-0.0691
0.3618
AR
2
-0.2941
0.1450
MA
1
0.5637
0.3737
MA
2
0.0840
0.3266

T
-0.19
-2.03
1.51
0.26

Differencing: 1 regular difference
Number of observations: Original series 100, after differencing 99
Residuals:
SS = 93.6368 (backforecasts excluded)
MS = 0.9857 DF = 95
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
11.0(DF= 8)
23.4(DF=20)
30.1(DF=32)

48
36.5(DF=44)

:‫وﻧﻠﺨﺺ ذﻟﻚ ﺑﺎﻟﺠﺪول اﻟﺘﺎﻟﻲ‬
Model
σˆ
__________ ________
ARI (1,1)
1.359
ARI ( 2,1)
1.080
IMA (1,1)
1.035
IMA (1, 2 )
1.023
ARIMA (1,1,1)
1.035
ARIMA ( 2,1,1)
1.398
ARIMA (1,1, 2 )
1.013
ARIMA ( 2,1, 2 )
0.986
2

m
___
2
3
2
3
3
4
4
5

AIC
_________
34.368
13.619
7.4057
8.2706
9.4057
41.169
9.2689
8.5741

min AIC ( m ) = 7.406
m

.‫ ﻳﺘﺮك ﻟﻠﻄﺎﻟﺐ آﺘﻤﺮﻳﻦ ﻓﺤﺺ اﻟﺒﻮاﻗﻲ وﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‬. IMA(١،١) ‫أي ان أﻓﻀﻞ ﻧﻤﻮذج هﻮ‬

١٠٤

١٠٥

‫اﻟﻔﺼﻞ اﻟﺴﺎدس‬
‫ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﺘﻜﺎﻣﻠﻲ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻤﻮﺳﻤﻴﺔ‬

‫‪Seasonal Autoregressive Integrated Moving‬‬
‫‪Average Models‬‬
‫اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ ﺗﻌﻄﻲ اﻧﻤﺎط ﻣﺘﺸﺎﺑﻬﺔ ﺗﺘﻜﺮر ﻋﻠﻰ ﻓﺘﺮات زﻣﻨﻴﺔ ﻣﺘﺴﺎوﻳﺔ اﻟﺒﻌﺪ ﻣﺜﻞ‬
‫ان ﻳﺘﻜﺮر اﻟﻨﻤﻂ آﻞ ارﺑﻌﺔ وﻋﺸﺮون ﺳﺎﻋﺔ او آﻞ ﺳﺒﻌﺔ اﻳﺎم او آﻞ ﺷﻬﺮ او ﺛﻼﺛﺔ اﺷﻬﺮ او ﺳﻨﺔ‪.‬‬
‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ ﺗﺒﻴﻦ ﻣﺜﻞ هﺬﻩ اﻟﻤﺘﺴﻠﺴﻼت‬
‫‪S e a s o n a l T im e S e r ie s‬‬

‫‪70‬‬

‫)‪z(t‬‬
‫‪60‬‬

‫‪50‬‬

‫‪150‬‬

‫‪50‬‬

‫‪100‬‬

‫‪In d e x‬‬

‫‪S e a s o n a l T im e S e r ie s‬‬
‫‪1000‬‬

‫‪900‬‬

‫)‪z(t‬‬

‫‪800‬‬

‫‪700‬‬

‫‪600‬‬

‫‪150‬‬

‫‪100‬‬

‫‪In d e x‬‬

‫‪50‬‬

‫ﻓﻲ هﺬا اﻟﻔﺼﻞ ﺳﻮف ﻧﺴﺘﻌﺮض ﺧﻮاص اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ وﻃﺮق ﻧﻤﺬﺟﺘﻬﺎ ﺑﻮاﺳﻄﺔ‬
‫ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﺘﻜﺎﻣﻠﻲ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪ SARIMA(p,d,q)(P,D,Q)s‬ﻓﻤﺜﻼ‬
‫اﻟﻨﻤﻮذج ‪ SARIMA(٠،١،١)(١،١،٠)١٢‬ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬
‫) ‪WN ( 0, σ 2‬‬

‫‪١٠٦‬‬

‫‪at‬‬

‫‪(1 − Φ B ) (1 − B ) z = (1 − θ B ) a ,‬‬
‫‪s‬‬

‫‪t‬‬

‫‪1‬‬

‫‪t‬‬

‫‪1‬‬

‫وﺑﺸﻜﻞ ﻋﺎم ﻓﺈن ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﺘﻜﺎﻣﻠﻲ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﺑﺎﻟﺪرﺟﺔ ‪(p,d,q)(P,D,Q)s‬‬
‫‪ SARIMA(p,d,q)(P,D,Q)s‬ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪φ p ( B ) Φ P ( B s ) (1 − B ) (1 − B s ) zt = δ + θ q ( B ) ΘQ ( B s ) at , at WN ( 0, σ 2‬‬
‫‪D‬‬

‫‪d‬‬

‫ﺣﻴﺚ ) ‪ φ p ( B‬و ) ‪ θ q ( B‬ﻋﻤﺎل اﻹﻧﺤﺪار اﻟﺬاﺗﻲ واﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻏﻴﺮ اﻟﻤﻮﺳﻤﻴﺔ واﻟﺘﻲ ﻣﺮت‬
‫ﻋﻠﻴﻨﺎ ﺳﺎﺑﻘﺎ و ‪ Φ P ( B s ) = 1 + Φ1B s + Φ 2 B 2 s + L + Φ P B Ps‬ﻋﺎﻣﻞ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﻤﻮﺳﻤﻲ و‬
‫‪ ΘQ ( B s ) = 1 + Θ1B s + Θ2 B 2 s + L + ΘQ B Qs‬ﻋﺎﻣﻞ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻤﻮﺳﻤﻲ وﻳﺴﻤﻰ هﺬا‬
‫ﺑﺎﻟﻨﻤﻮذج اﻟﻤﻮﺳﻤﻲ اﻟﺘﻀﺎﻋﻔﻲ ‪.Multiplicative Seasonal Models‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬ﻓﻲ ﺟﻤﻴﻊ اﻟﻨﻤﺎذج اﻟﻘﺎدﻣﺔ ﺳﻴﻜﻮن ﻣﻔﻬﻮﻣﺎ ﺿﻤﻨﻴﺎ أن ) ‪at WN ( 0, σ 2‬‬

‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ‪:‬‬
‫‪D‬‬
‫‪d‬‬
‫ﻓﻲ اﻹﺷﺘﻘﺎﻗﺎت اﻟﺘﺎﻟﻴﺔ ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻮﺳﻤﻴﺔ اﻟﻤﺴﺘﻘﺮة ‪wt = (1 − B ) (1 − B s ) zt‬‬
‫واﻟﺘﻲ ﺗﺘﺒﻊ اﻟﻨﻤﻮذج ‪SARMA(p,q)(P,Q)s‬‬
‫) ‪WN ( 0, σ 2‬‬

‫‪φ p ( B ) Φ P ( B ) wt = δ + θ q ( B ) ΘQ ( B ) at , at‬‬
‫‪s‬‬

‫‪s‬‬

‫ﺳﻮف ﻧﺸﺘﻖ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻨﻤﻮذج اﻟﻤﻮﺳﻤﻲ اﻟﺘﻀﺎﻋﻔﻲ ‪SARMA(٠،١)(١،١)١٢‬‬
‫) ‪WN ( 0, σ 2‬‬

‫‪wt = Φ wt −12 + at − θ at −1 − Θat −12 + θ Θat −13 , at‬‬

‫ﺑﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻓﻲ ‪ wt‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬
‫‪γ 0 = Φ γ 12 + σ + θ σ − Θ ( Φ − Θ ) σ 2 + θ Θ ( −Φ θ + θ Θ ) σ 2‬‬
‫‪2‬‬

‫‪2‬‬

‫‪2‬‬

‫⎦⎤) ‪=Φ γ 12 + σ 2 ⎡⎣(1 + θ 2 ) + Θ ( Φ − Θ ) (1 + θ 2‬‬
‫⎦⎤) ‪=Φ γ 12 + σ 2 (1 + θ 2 ) ⎡⎣1 + Θ ( Φ − Θ‬‬

‫وﺑﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻓﻲ ‪ wt −12‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬
‫‪2‬‬

‫‪γ 12 = Φ γ 0 − Θσ + θ Θ ( −θ ) σ‬‬
‫‪2‬‬

‫) ‪=Φ γ 0 − Θσ 2 (1 + θ 2‬‬

‫وﺑﺤﻞ اﻟﻌﻼﻗﺘﻴﻦ اﻟﺴﺎﺑﻘﺘﻴﻦ ﻧﺠﺪ‬
‫‪1 + Θ − 2ΦΘ‬‬
‫‪1 − Φ2‬‬
‫‪2‬‬
‫⎡‬
‫⎤ ) ‪Φ (Θ − Φ‬‬
‫‪2‬‬
‫‪2‬‬
‫‪= σ (1 + θ ) ⎢Φ − Θ +‬‬
‫⎥‬
‫⎦⎥ ‪1 − Φ 2‬‬
‫⎣⎢‬
‫‪2‬‬

‫أﻳﻀﺎ‬

‫) ‪γ 0 = σ 2 (1 + θ 2‬‬
‫‪γ 12‬‬

‫) ‪γ 1 = E ( wt wt −1‬‬
‫) ‪=Φ γ 11 − θσ 2 − ΘE ( at −12 wt −1 ) + θ ΘE ( at −13wt −1‬‬
‫‪=Φ γ 11 − θσ 2 + θ Θ ( Φ − Θ ) σ 2‬‬

‫و‬
‫‪2‬‬

‫وﺑﺤﻞ اﻟﻌﻼﻗﺘﻴﻦ اﻟﺴﺎﺑﻘﺘﻴﻦ ﻧﺠﺪ‬

‫‪١٠٧‬‬

‫‪γ 11 = E ( wt wt −11 ) = Φγ 1 + Θθσ‬‬

⎡ ( Θ − Φ )2 ⎤
γ 1 = −θσ ⎢1 +

1 − Φ 2 ⎦⎥
⎣⎢
2

γ 11 = θσ 2 ⎢ Θ − Φ −
⎣⎢

2
Φ (Θ − Φ ) ⎤

1 − Φ 2 ⎦⎥

‫وﺑﻨﻔﺲ اﻟﻄﺮﻳﻘﺔ ﻳﻤﻜﻦ إﺛﺒﺎت أن‬

γ 2 = γ 3 = L = γ 10 = 0
γ 13 = γ 11
γ k = Φγ k −12 , k > 13

‫وﻣﻦ ﺟﻤﻴﻊ اﻟﻌﻼﻗﺎت اﻟﺴﺎﺑﻘﺔ ﻧﻮﺟﺪ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬
k =0
⎧ 1,

θ
⎪ −
,
k =1
2
⎪ 1+θ
⎪ 0,
k = 2,...,10

γ
⎪ θ ( Θ − Φ )(1 − ΦΘ )
ρk = k = ⎨
, k = 11
γ 0 ⎪1 + θ 2 1 + Θ2 − 2ΦΘ
⎪ ( Θ − Φ )(1 − ΦΘ )
k = 12
,
⎪−
2
1
+
Θ

2
ΦΘ

k = 13
⎪ ρ11 ,
⎪ Φρ ,
k > 13
k −12

:‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ‬

wt = (1 − ΘB s ) at SARIMA(٠,d,٠)(٠,D,١)s ‫ ﻧﻤﻮذج‬-١

k =0
⎧ 1,
⎪⎪
Θ
,
k=s
ρk = ⎨−
2
1
+
Θ

otherwise
⎪⎩ 0,

(1 − Φ B ) w
s

t

= at

SARIMA(٠,d,٠)(١,D,١)s ‫ ﻧﻤﻮذج‬-٢

k =0
⎧1,
⎪ ks
ρk = ⎨Φ , k = s, 2 s,...
⎪ 0,
otherwise

wt = (1 − θ B ) (1 − ΘB s ) at

١٠٨

SARIMA(٠,d,١)(٠,D,١)s ‫ ﻧﻤﻮذج‬-٣

k =0
⎧ 1,

θ
⎪−
,
k =1
2
⎪ 1+θ

θΘ
, k = s −1
⎪⎪
2
2
ρ k = ⎨ (1 + θ )(1 + Θ )

Θ
⎪−
,
k=s
2
⎪ 1+ Θ
⎪ ρ s−1 ,
k = s +1

otherwise
⎪⎩ 0,

(1 − ΦB ) w = (1 − ΘB ) a
s

s

t

t

k =0
⎧ 1,

⎪ ( Θ − Φ )(1 − ΦΘ ) k s−1
ρk = ⎨−
Φ , k = s, 2 s,...
2
⎪ 1 + Θ − 2ΦΘ
otherwise
⎪⎩ 0,
(1 − ΦB s ) wt = (1 − θ B ) at

SARIMA(٠,d,٠)(١,D,١)s ‫ ﻧﻤﻮذج‬-٤

SARIMA(٠,d,١)(١,D,٠)s ‫ ﻧﻤﻮذج‬-٥

k =0
⎧ 1,

θ
⎪−
, k =1
2
⎪ 1+θ
⎪ 0,
k = 2,..., s − 2

ρk = ⎨ θ Φ

, k = s −1
⎪ 1+θ 2

k=s
⎪ Φ,
⎪ ρ s−1 ,
k = s +1

k > s +1
⎪⎩ Φ ρ k − s ,

wt = (1 − θ1B − θ 2 B 2 )(1 − ΘB12 ) at

١٠٩

SARIMA(٠,d,٢)(٠,D,١)s ‫ ﻧﻤﻮذج‬-٦

‫‪k =0‬‬
‫‪k =1‬‬
‫‪k =2‬‬
‫‪k = s−2‬‬
‫‪k = s −1‬‬
‫‪k=s‬‬
‫‪k = s +1‬‬
‫‪k = s+2‬‬
‫‪otherwise‬‬

‫‪⎧ 1,‬‬
‫⎪‬
‫‪⎪ − θ1 (1 − θ 2 ) ,‬‬
‫‪⎪ 1 + θ12 + θ 22‬‬
‫⎪‬
‫‪θ2‬‬
‫‪⎪−‬‬
‫‪,‬‬
‫‪⎪ 1 + θ12 + θ 22‬‬
‫⎪‬
‫‪θ 2Θ‬‬
‫⎪‬
‫‪,‬‬
‫‪2‬‬
‫) ‪⎪⎪ (1 + θ1 + θ 22 )(1 + Θ2‬‬
‫⎨ = ‪ρk‬‬
‫) ‪θ1Θ (1 − θ 2‬‬
‫⎪‬
‫‪,‬‬
‫) ‪⎪ (1 + θ 2 + θ 2 )(1 + Θ2‬‬
‫‪1‬‬
‫‪2‬‬
‫⎪‬
‫⎪‬
‫‪Θ‬‬
‫‪,‬‬
‫‪⎪−‬‬
‫‪2‬‬
‫‪+‬‬
‫‪Θ‬‬
‫‪1‬‬
‫⎪‬
‫‪⎪ ρ s −1 ,‬‬
‫‪⎪ ρ ,‬‬
‫‪⎪ s −2‬‬
‫‪⎪⎩ 0,‬‬

‫ﺳﻮف ﻧﺴﺘﻌﺮض ﺑﻌﺾ اﻟﺮﺳﻮﻣﺎت ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ ﻷﻋﻄﺎء‬
‫ﻓﻜﺮة ﻋﻦ أﺷﻜﺎﻟﻬﺎ‪.‬‬
‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ ﻟﻨﻤﻮذج ‪: SARIMA(٠,d,١)(١,D,٠)١٢‬‬
‫ﺷﻜﻞ )‪(١‬‬

‫‪Φ = 0.6, θ = 0.5‬‬
‫‪A ( 0 ,d ,1 ) ( 1 ,D ,0 ) 1 2‬‬

‫‪o f S A R I M‬‬

‫‪A C F‬‬
‫‪0 .5‬‬

‫‪C1‬‬

‫‪0 .0‬‬

‫‪-0 .5‬‬

‫‪5 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪3 0‬‬

‫‪0‬‬

‫‪1 0‬‬

‫‪L a g‬‬

‫ﺷﻜﻞ )‪( ٢‬‬

‫‪Φ = 0.6, θ = −0.5‬‬
‫‪A ( 0 ,d ,1 ) ( 1 ,D ,0 ) 1 2‬‬

‫‪o f S A R I M‬‬

‫‪A C F‬‬
‫‪0 .7‬‬
‫‪0 .6‬‬
‫‪0 .5‬‬

‫‪0 .3‬‬
‫‪0 .2‬‬
‫‪0 .1‬‬
‫‪0 .0‬‬

‫‪5 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪3 0‬‬

‫‪L a g‬‬

‫‪١١٠‬‬

‫‪1 0‬‬

‫‪0‬‬

‫‪C1‬‬

‫‪0 .4‬‬

‫ﺷﻜﻞ )‪(٣‬‬

‫‪Φ = −0.6, θ = 0.5‬‬
‫‪A C F o f S A R IM A (0 ,d ,1 )(1 ,D ,0 )1 2‬‬
‫‪0 .5‬‬

‫‪C1‬‬

‫‪0 .0‬‬

‫‪-0 .5‬‬

‫‪50‬‬

‫‪40‬‬

‫‪20‬‬

‫‪30‬‬

‫‪10‬‬

‫‪0‬‬

‫‪Lag‬‬

‫ﺷﻜﻞ )‪(٤‬‬

‫‪Φ = −0.6, θ = −0.5‬‬
‫‪A C F o f S A R IM A (0 ,d ,1 )(1 ,D ,0 )1 2‬‬

‫‪0 .5‬‬

‫‪C1‬‬

‫‪0 .0‬‬

‫‪-0 .5‬‬

‫‪5 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪3 0‬‬

‫‪1 0‬‬

‫‪0‬‬

‫‪L ag‬‬

‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻨﻤﻮذج اﻟﻤﻮﺳﻤﻲ اﻟﺘﻀﺎﻋﻔﻲ‪:‬‬
‫ﻣﻦ اﻟﺼﻌﻮﺑﺔ إﺷﺘﻘﺎق وﺗﻔﺴﻴﺮ أﻧﻤﺎط داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ اﻟﺘﻀﺎﻋﻔﻴﺔ‬
‫وﻟﻜﻨﻬﺎ وﺑﺸﻜﻞ ﻋﺎم ﻓﺈن أﺟﺰاء اﻟﻨﻤﻮذج اﻟﻤﻮﺳﻤﻴﺔ وﻏﻴﺮ اﻟﻤﻮﺳﻤﻴﺔ واﻟﺘﻲ ﺗﻨﻤﺬج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‬
‫ﺗﻌﻄﻲ ﺗﺨﺎﻣﺪات اﺳﻴﺔ وﺗﺨﺎﻣﺪات ﺟﻴﺒﻴﺔ ﻋﻨﺪ اﻟﺘﺨﻠﻔﺎت اﻟﻤﻮﺳﻤﻴﺔ وﻏﻴﺮاﻟﻤﻮﺳﻤﻴﺔ وﻓﻲ اﻟﻨﻤﺎذج اﻟﺘﻲ‬
‫ﺗﺤﻮي إﻧﺤﺪار ذاﺗﻲ ﻓﺈن اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ ﺗﻌﻄﻲ ﻗﻄﻌﺎ ‪. cut off‬‬
‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ ﻹﻋﻄﺎء ﻓﻜﺮة ﻋﻦ ﺑﻌﺾ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ‪:‬‬
‫‪ -١‬ﺷﻜﻞ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج ‪wt = (1 − ΘB12 ) at‬‬
‫ا( ‪Θ = 0.6‬‬

‫‪١١١‬‬

A C F o f S A R IM A (0 ,d ,0 )(0 ,D ,1 )1 2
0 .0
-0 .1

C1

-0 .2
-0 .3
-0 .4
-0 .5
-0 .6

0

1 0

2 0

3 0

4 0

5 0

L ag

Θ = −0.6 (‫ب‬
A C F o f S A R IM A (0 ,d ,0 )(0 ,D ,1 )1 2
0 .6
0 .5
0 .4

C1

0 .3
0 .2
0 .1
0 .0
-0 .1
-0 .2
0

1 0

2 0

3 0

4 0

5 0

L ag

(1 − Φ B ) w
12

t

= at ‫ ﺷﻜﻞ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج‬-٢

Φ = 0.6 (‫ا‬
A C F o f S A R IM A (0 ,d ,1 )(0 ,D ,0 )1 2
0 .6
0 .5

C1

0 .4
0 .3
0 .2
0 .1
0 .0

0

1 0

2 0

3 0

4 0

5 0

L a g

Φ = −0.6 (‫ب‬

١١٢

‫‪A C F o f S A R IM A (0 ,d ,1 )(0 ,D ,0 )1 2‬‬
‫‪0 .0‬‬
‫‪-0 .1‬‬
‫‪-0 .2‬‬

‫‪C1‬‬

‫‪-0 .3‬‬
‫‪-0 .4‬‬
‫‪-0 .5‬‬
‫‪-0 .6‬‬

‫‪5 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪3 0‬‬

‫‪0‬‬

‫‪1 0‬‬

‫‪L ag‬‬

‫ﺃﻤﺜﻠﺔ‪ :‬ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ )ﻓﻲ ﺠﻤﻴﻊ ﺍﻷﻤﺜﻠﺔ ﺍﻟﺘﺎﻟﻴﺔ ﺇﻗﺭﺃ ﺴﻁﺭﺍ ﺒﺴﻁﺭ(‬
‫‪54.9‬‬
‫‪58.2‬‬
‫‪68.2‬‬
‫‪54.9‬‬
‫‪53.2‬‬
‫‪57.3‬‬
‫‪70.4‬‬
‫‪53.4‬‬
‫‪51.7‬‬
‫‪57.1‬‬
‫‪75.5‬‬

‫‪54.9‬‬
‫‪54.3‬‬
‫‪70.1‬‬
‫‪57.5‬‬
‫‪52.8‬‬
‫‪52.7‬‬
‫‪69.9‬‬
‫‪55.3‬‬
‫‪51.5‬‬
‫‪53.9‬‬
‫‪73.3‬‬

‫‪55.3‬‬
‫‪53.4‬‬
‫‪67.5‬‬
‫‪61.2‬‬
‫‪52.8‬‬
‫‪51.6‬‬
‫‪61.0‬‬
‫‪58.2‬‬
‫‪52.3‬‬
‫‪53.5‬‬
‫‪68.1‬‬
‫‪62.2‬‬

‫‪56.9‬‬
‫‪53.0‬‬
‫‪58.0‬‬
‫‪69.3‬‬
‫‪54.5‬‬
‫‪52.4‬‬
‫‪52.7‬‬
‫‪66.9‬‬
‫‪53.6‬‬
‫‪53.5‬‬
‫‪58.1‬‬
‫‪74.8‬‬

‫‪57.4‬‬
‫‪52.8‬‬
‫‪54.2‬‬
‫‪69.8‬‬
‫‪56.4‬‬
‫‪52.1‬‬
‫‪53.9‬‬
‫‪70.7‬‬
‫‪55.3‬‬
‫‪53.1‬‬
‫‪54.8‬‬
‫‪76.4‬‬

‫‪61.5‬‬
‫‪53.3‬‬
‫‪53.2‬‬
‫‪66.1‬‬
‫‪59.3‬‬
‫‪52.6‬‬
‫‪53.3‬‬
‫‪65.3‬‬
‫‪58.5‬‬
‫‪53.3‬‬
‫‪54.3‬‬
‫‪70.8‬‬

‫‪72.7‬‬
‫‪54.4‬‬
‫‪53.0‬‬
‫‪56.1‬‬
‫‪68.7‬‬
‫‪53.9‬‬
‫‪53.5‬‬
‫‪56.5‬‬
‫‪69.3‬‬
‫‪53.9‬‬
‫‪54.6‬‬
‫‪60.6‬‬

‫‪72.2‬‬
‫‪56.0‬‬
‫‪53.0‬‬
‫‪53.6‬‬
‫‪70.0‬‬
‫‪56.4‬‬
‫‪53.4‬‬
‫‪53.4‬‬
‫‪69.6‬‬
‫‪55.6‬‬
‫‪54.2‬‬
‫‪56.4‬‬

‫‪71.5‬‬
‫‪60.0‬‬
‫‪53.4‬‬
‫‪53.0‬‬
‫‪67.9‬‬
‫‪61.7‬‬
‫‪53.6‬‬
‫‪52.5‬‬
‫‪64.2‬‬
‫‪60.1‬‬
‫‪54.8‬‬
‫‪55.6‬‬

‫‪59.1‬‬
‫‪71.0‬‬
‫‪54.6‬‬
‫‪52.8‬‬
‫‪58.7‬‬
‫‪68.3‬‬
‫‪55.1‬‬
‫‪53.2‬‬
‫‪55.5‬‬
‫‪68.9‬‬
‫‪55.8‬‬
‫‪55.0‬‬

‫‪57.2‬‬
‫‪70.6‬‬
‫‪55.6‬‬
‫‪52.6‬‬
‫‪55.4‬‬
‫‪67.9‬‬
‫‪57.3‬‬
‫‪53.0‬‬
‫‪53.3‬‬
‫‪68.8‬‬
‫‪57.9‬‬
‫‪54.7‬‬

‫‪56.3‬‬
‫‪68.2‬‬
‫‪59.4‬‬
‫‪52.9‬‬
‫‪52.9‬‬
‫‪65.3‬‬
‫‪61.9‬‬
‫‪53.5‬‬
‫‪52.4‬‬
‫‪63.6‬‬
‫‪62.6‬‬
‫‪55.8‬‬

‫‪55.8‬‬
‫‪57.7‬‬
‫‪69.8‬‬
‫‪54.0‬‬
‫‪53.0‬‬
‫‪58.2‬‬
‫‪69.9‬‬
‫‪54.3‬‬
‫‪51.5‬‬
‫‪57.1‬‬
‫‪70.3‬‬
‫‪57.7‬‬

‫‪55.7‬‬
‫‪54.6‬‬
‫‪71.0‬‬
‫‪54.9‬‬
‫‪52.7‬‬
‫‪55.8‬‬
‫‪71.5‬‬
‫‪56.3‬‬
‫‪51.5‬‬
‫‪52.2‬‬
‫‪69.4‬‬
‫‪60.5‬‬

‫ﺷﻜﻞ اﻟﻤﺘﺴﻠﺴﻠﺔ هﻮ‬

‫‪7 0‬‬

‫)‪z(t‬‬

‫‪6 0‬‬

‫‪5 0‬‬

‫‪1 5 0‬‬

‫‪1 0 0‬‬

‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬

‫‪١١٣‬‬

‫‪5 0‬‬

‫‪In d e x‬‬

‫)‪z(t‬‬
‫‪56.3‬‬
‫‪54.9‬‬
‫‪67.4‬‬
‫‪56.6‬‬
‫‪53.4‬‬
‫‪55.3‬‬
‫‪65.1‬‬
‫‪59.4‬‬
‫‪52.1‬‬
‫‪51.5‬‬
‫‪64.7‬‬
‫‪66.4‬‬

Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r z ( t )
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

2
Lag
1
2
3
4
5
6
7
8
9
10
11
12

C o rr

12

T

LBQ

Lag

C o rr

0 .7 7 1 0 .3 1
2 .9 3
0 .3 3
-0 .1 1 -0 .9 9
-0 .3 8 -3 .2 8
-0 .5 2 -4 .2 2
-0 .5 6 -4 .1 6
-0 .5 2 -3 .5 0
-0 .3 8 -2 .3 9
-0 .1 2 -0 .7 1
1 .7 6
0 .2 9
4 .1 5
0 .6 9
4 .8 6
0 .8 8

1 0 8 .1 0
1 2 7 .3 4
1 2 9 .7 4
1 5 6 .8 3
2 0 7 .2 4
2 6 6 .5 5
3 1 6 .8 0
3 4 3 .7 1
3 4 6 .2 5
3 6 2 .0 8
4 5 3 .5 8
6 0 3 .8 7

13
14
15
16
17
18
19
20
21
22
23
24

0 .6 8
0 .2 8
-0 .1 2
-0 .3 6
-0 .4 9
-0 .5 3
-0 .4 9
-0 .3 7
-0 .1 4
0 .2 3
0 .6 1
0 .7 9

22
T

32

42

LBQ

Lag

C o rr

T

LBQ

Lag

C o rr

T

LBQ

3 .3 4 6 9 4 .2 8
1 .3 0 7 0 9 .8 2
-0 .5 3 7 1 2 .4 3
-1 .6 5 7 3 8 .3 0
-2 .2 1 7 8 6 .2 8
-2 .3 3 8 4 3 .0 4
-2 .1 0 8 9 2 .3 6
-1 .5 4 9 2 0 .3 9
-0 .5 6 9 2 4 .2 4
0 .9 5 9 3 5 .3 7
2 .4 6 1 0 1 1 .3 1
3 .1 0 1 1 4 0 .9 0

25
26
27
28
29
30
31
32
33
34
35
36

0 .6 1
0 .2 5
-0 .1 1
-0 .3 4
-0 .4 7
-0 .5 1
-0 .4 8
-0 .3 6
-0 .1 4
0 .1 9
0 .5 4
0 .7 2

2 .3 0
0 .9 1
-0 .4 2
-1 .2 4
-1 .6 7
-1 .8 0
-1 .6 5
-1 .2 3
-0 .4 9
0 .6 6
1 .8 2
2 .3 7

1 2 2 0 .0 7
1 2 3 3 .1 8
1 2 3 5 .9 8
1 2 6 1 .2 9
1 3 0 8 .3 2
1 3 6 4 .7 2
1 4 1 4 .0 8
1 4 4 2 .7 1
1 4 4 7 .3 4
1 4 5 5 .7 9
1 5 2 0 .8 9
1 6 3 6 .4 9

37
38
39
40
41
42
43
44

0 .5 7
0 .2 3
-0 .1 1
-0 .3 3
-0 .4 4
-0 .4 8
-0 .4 5
-0 .3 4

1 .8 2
0 .7 3
-0 .3 4
-1 .0 2
-1 .3 8
-1 .4 8
-1 .3 7
-1 .0 4

1 7 0 9 .3 6
1 7 2 1 .7 0
1 7 2 4 .4 3
1 7 4 9 .0 6
1 7 9 5 .0 0
1 8 4 9 .4 1
1 8 9 7 .0 5
1 9 2 5 .3 0

‫واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬
Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r z ( t )
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
- 0 .2
- 0 .4
- 0 .6
- 0 .8
- 1 .0

2

12

22

32

42

Lag

P AC

T

Lag

P AC

T

Lag

P AC

T

Lag

P AC

T

1
2
3
4
5
6
7
8
9
10
11
12

0 .7 7
-0 .6 8
-0 .0 6
0 .0 1
-0 .4 2
-0 .1 8
-0 .1 1
-0 .2 2
0 .1 8
0 .5 7
0 .2 6
0 .1 6

1 0 .3 1
- 9 .0 1
- 0 .8 2
0 .0 7
- 5 .5 5
- 2 .4 5
- 1 .4 8
- 2 .9 6
2 .3 9
7 .6 5
3 .5 0
2 .1 6

13
14
15
16
17
18
19
20
21
22
23
24

- 0 .4 2
0 .2 9
0 .0 1
- 0 .0 7
0 .0 7
- 0 .0 1
- 0 .0 9
- 0 .0 5
- 0 .0 8
0 .1 0
- 0 .0 3
0 .0 3

- 5 .6 2
3 .8 9
0 .1 6
- 0 .9 2
0 .9 5
- 0 .1 1
- 1 .1 5
- 0 .7 2
- 1 .1 1
1 .3 8
- 0 .3 5
0 .3 7

25
26
27
28
29
30
31
32
33
34
35
36

- 0 .1 3
- 0 .0 0
0 .0 4
- 0 .1 0
- 0 .0 3
0 .0 1
0 .0 1
- 0 .0 4
- 0 .0 4
0 .0 1
- 0 .0 2
0 .0 7

-1 .7 5
-0 .0 5
0 .5 0
-1 .3 0
-0 .3 4
0 .1 9
0 .0 7
-0 .5 5
-0 .5 2
0 .1 5
-0 .3 2
0 .9 0

37
38
39
40
41
42
43
44

-0 .0 6
-0 .0 5
0 .0 1
0 .0 0
-0 .0 5
0 .0 5
-0 .0 5
-0 .0 3

-0 .7 4
-0 .6 2
0 .1 7
0 .0 0
-0 .6 3
0 .6 6
-0 .6 4
-0 .4 2

.‫ﻧﻼﺣﻆ اﻷﻧﻤﺎط اﻟﻤﻮﺳﻤﻴﺔ واﺿﺤﺔ ﻓﻲ اﻷﺷﻜﺎل اﻟﺴﺎﺑﻘﺔ‬
‫ﻤﺜﺎل ﺁﺨﺭ‬
z(t)
589
673
678
621
713
796
801
747
826
898
908
827

561
742
639
602
667
858
764
711
799
957
867
797

640
716
604
635
762
826
725
751
890
924
815
843

656
660
611
677
784
783
723
804
900
881
812

727
617
594
635
837
740
690
756
961
837
773

697
583
634
736
817
701
734
860
935
784
813

640
587
658
755
767
706
750
878
894
791
834

599
565
622
811
722
677
707
942
855
760
782

568
598
709
798
681
711
807
913
809
802
892

577
628
722
735
687
734
824
869
810
828
903

553
618
782
697
660
690
886
834
766
778
966

582
688
756
661
698
785
859
790
805
889
937

600
705
702
667
717
805
819
800
821
902
896

566
770
653
645
696
871
783
763
773
969
858

653
736
615
688
775
845
740
800
883
947
817

‫ﺷﻜﻞ اﻟﻤﺘﺴﻠﺴﻠﺔ‬

١١٤

1000

z(t)

900

800

700

600

50

In d e x

100

150

‫اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬
Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r z ( t )
1
0
0
0
0
0
-0
-0
-0
-0
-1

.0
.8
.6
.4
.2
.0
.2
.4
.6
.8
.0

2
Lag
1
2
3
4
5
6
7
8
9
10
11
12

C o rr
0
0
0
0
0
0
0
0
0
0
0
0

.8
.7
.6
.4
.4
.3
.4
.4
.5
.6
.7
.8

T

12
LB Q

9 1 1 .5 6 1 3 5 .9 4
6 .2 7 2 4 0 .1 3
8
4 .1 2 3 0 6 .7 2
2
2 .9 5 3 4 7 .9 7
9
2 .4 7 3 8 0 .0 9
3
2 .1 0 4 0 5 .0 2
8
2 .2 5 4 3 5 .5 4
1
2 .4 0 4 7 2 .3 7
5
2 .8 7 5 2 9 .0 7
6
3 .3 4 6 1 4 .2 7
9
3 .5 2 7 2 1 .7 2
7
3 .6 1 8 5 2 .4 1
4

L ag
1
1
1
1
1
1
1
2
2
2
2
2

3
4
5
6
7
8
9
0
1
2
3
4

22

C o rr
0
0
0
0
0
0
0
0
0
0
0
0

.7
.6
.4
.3
.3
.2
.2
.3
.4
.5
.6
.6

4
4
9
6
1
5
9
2
2
3
0
7

T
2
2
1
1
1
0
1
1
1
1
2
2

.9
.4
.7
.3
.0
.9
.0
.1
.4
.8
.0
.2

6
11
91
11
91
01
11
21
41
11
31
11

LB Q
9
0
0
0
1
1
1
1
1
2
3
4

5
3
7
9
1
2
4
6
9
5
2
1

4
0
4
9
7
9
5
5
9
3
5
5

.6
.0
.8
.6
.3
.7
.5
.4
.1
.8
.5
.2

8
9
5
8
9
6
9
2
3
1
1
9

L ag
2
2
2
2
2
3
3
3
3
3
3
3

5
6
7
8
9
0
1
2
3
4
5
6

32

C o rr
0
0
0
0
0
0
0
0
0
0
0
0

.5
.4
.3
.2
.1
.1
.1
.2
.2
.3
.4
.5

8
9
5
4
9
4
7
0
8
8
5
2

T
1
1
1
0
0
0
0
0
0
1
1
1

.8
.5
.0
.7
.5
.4
.5
.6
.8
.1
.3
.5

61
21
91
31
71
31
11
01
51
51
51
31

LB Q
4
5
5
5
5
5
5
5
6
6
6
7

8
3
5
6
7
7
8
9
1
4
8
4

3
0
6
8
5
9
5
3
0
1
5
2

.1
.9
.4
.2
.4
.5
.5
.6
.3
.5
.2
.9

1
9
5
0
4
5
2
2
5
7
7
5

L ag
3
3
3
4
4
4

7
8
9
0
1
2

42
C o rr
0
0
0
0
0
0

.4
.3
.2
.1
.0
.0

3
5
2
2
6
2

T
1
1
0
0
0
0

.2
.0
.6
.3
.1
.0

71
01
41
31
81
51

LB Q
7
8
8
8
8
8

8
1
2
2
2
2

4
0
1
4
5
5

.1
.6
.6
.6
.5
.6

7
0
8
7
6
2

‫واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬
Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r z ( t )
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

2

12

22

32

42

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

1
2
3
4
5
6
7
8
9
10
11
12

0 .8 9
-0 .0 8
-0 .2 8
0 .0 3
0 .3 5
-0 .0 8
0 .2 8
0 .0 9
0 .4 0
0 .3 0
0 .0 6
0 .2 2

1 1 .5 6
-1 .0 6
-3 .6 5
0 .4 2
4 .5 4
-1 .0 7
3 .6 7
1 .1 9
5 .1 7
3 .9 5
0 .8 1
2 .8 8

13
14
15
16
17
18
19
20
21
22
23
24

-0 .6 3
-0 .0 2
0 .0 7
-0 .0 4
-0 .0 9
-0 .0 4
-0 .0 5
0 .0 3
0 .0 4
0 .0 5
0 .0 5
0 .0 5

-8 .1 9
-0 .2 1
0 .9 5
-0 .5 2
-1 .1 2
-0 .4 9
-0 .6 0
0 .3 8
0 .4 6
0 .6 7
0 .6 0
0 .5 9

25
26
27
28
29
30
31
32
33
34
35
36

-0 .1 8
0 .0 8
0 .0 6
-0 .0 3
-0 .0 4
0 .0 0
-0 .0 6
-0 .0 1
-0 .0 1
0 .0 3
0 .0 0
0 .0 1

-2 .3 6
1 .0 6
0 .7 3
-0 .4 4
-0 .4 7
0 .0 2
-0 .7 2
-0 .1 1
-0 .1 8
0 .3 8
0 .0 3
0 .0 9

37
38
39
40
41
42

-0 .1 1
-0 .0 2
0 .0 4
-0 .0 3
-0 .0 8
0 .0 1

-1 .3 7
-0 .2 2
0 .5 1
-0 .4 2
-1 .0 6
0 .0 8

١١٥

‫ﻣﺜﺎل ﺁﺧﺮ‬
z(t)
302
107
055
237
079
035
256

262
056
048
247
045
056
250

218
049
115
215
040
097
198

175
047
185
182
038
210
136

100
047
276
080
041
260
073

077
071
220
046
069
257
039

043
151
181
065
152
210
032

047
244
151
040
232
125
030

049
280
083
044
282
080
031

069
230
055
063
255
042
045

152
185
049
085
161
035

205
148
042
185
107
031

246
098
046
247
053
032

294
061
074
231
040
050

242
046
103
167
039
092

181
045
200
117
034
189

‫وﻟﻬﺎ اﻟﺸﻜﻞ‬

3 0 0

z(t)

2 0 0

1 0 0

0

1 0

In d e x

2 0

3 0

4 0

5 0

6 0

7 0

8 0

9 0

1 0 0

‫وداﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬
Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r z ( t )
1
0
0
0
0
0
-0
-0
-0
-0
-1

.0
.8
.6
.4
.2
.0
.2
.4
.6
.8
.0

5

1
2
3
4
5
6
7

C o rr
0
0
-0
-0
-0
-0
-0

.8
.4
.0
.4
.6
.7
.6

1
3
3
3
9
8
8

T
8
2
-0
-2
-4
-3
-3

.3
.8
.1
.7
.0
.9
.0

8
9
7
0
3
9
9

L B Q

1
1
2
2

7
9
9
1
6
3
8

2
2
2
3
6
5
9

.2
.3
.4
.1
.3
.1
.3

L a g

2
2
0
3
7
7
3

1
1
1
1
1

C o rr

T

8 -0 .4 3 -1 .7 8 3 1
9 -0 .0 4 -0 .1 8 3 1
0 0 .3 8 1 .5 4 3 2
1 0 .7 1 2 .8 2 3 8
2 0 .8 4 3 .1 0 4 7
3 0 .7 1 2 .4 2 5 3
4 0 .3 7 1 .2 0 5 5

L B Q
0
0
8
9
6
8
6

.6
.8
.2
.7
.2
.8
.1

L a g

6
8
2
2
2
4
0

1
1
1
1
1
2
2

5
6
7
8
9
0
1

2 5

C o rr
-0
-0
-0
-0
-0
-0
-0

.0
.3
.6
.6
.5
.3
.0

3
8
0
7
9
7
4

T
-0
-1
-1
-2
-1
-1
-0

.0
.1
.8
.0
.7
.0
.1

8
9
8
3
2
5
0

L B Q
5
5
6
6
7
7
7

5
7
2
7
2
4
4

6
4
0
8
4
2
2

.1
.0
.2
.3
.1
.2
.3

8
7
2
6
7
1
8

L a g
2
2
2
2
2

2
3
4
5
6

C o rr
0
0
0
0
0

.3
.6
.7
.5
.3

3
1
1
9
1

T
0
1
1
1
0

.9
.7
.9
.5
.8

2
1
3
5
0

L B Q
7
8
8
9
9

5
0
7
2
4

6
8
9
8
2

.9
.9
.4
.4
.4

8
7
2
1
6

‫وداﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬
P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r z ( t )
Partial Autocorrelation

L a g

1 5

1
0
0
0
0
0
-0
-0
-0
-0
-1

.0
.8
.6
.4
.2
.0
.2
.4
.6
.8
.0

5

L a g
1
2
3
4
5
6
7

P A C
0
-0
-0
-0
-0
-0
-0

.8
.7
.2
.2
.1
.3
.1

1
0
9
2
6
2
6

T
8
-7
-2
-2
-1
-3
-1

.3
.1
.9
.2
.6
.2
.6

8
7
8
8
1
9
0

1 5

L a g

1
1
1
1
1

P A C

8 -0 .0 3
9 0 .1 9
0 0 .2 7
1 0 .1 8
2 0 .0 7
3 -0 .1 0
4 -0 .1 5

-0
1
2
1
0
-0
-1

.3
.9
.7
.8
.7
.9
.5

T

L a g

1
1
6
2
4
9
2

1
1
1
1
1
2
2

١١٦

2 5

P A C

0 .1
5
0 .0
6
0 .1
7
0 .0
8
9 -0 .0
0 -0 .0
0 .0
1

9
8
1
4
2
3
4

1
0
1
0
-0
-0
0

.9
.8
.1
.3
.2
.2
.4

T

L a g

3
4
0
9
2
9
6

2
2
2
2
2

P A C

2 0 .0
3 0 .0
4 -0 .0
5 -0 .0
6 0 .1

1
7
7
2
0

T
0
0
-0
-0
1

.1
.7
.7
.2
.0

5
0
2
3
1

‫وآﻞ هﺬﻩ اﻟﻤﺘﻠﺴﻼت ﺗﺒﺪي اﻧﻤﺎط ﻣﻮﺳﻤﻴﺔ واﺿﺤﺔ‪.‬‬

‫إﺷﺘﻘﺎق دوال ﺗﻨﺒﺆ ﻟﺒﻌﺾ ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ اﻟﺘﻀﺎﻋﻔﻴﺔ‪:‬‬
‫ﺑﻤﺎ ان ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ هﻲ ﺣﺎﻟﺔ ﺧﺎﺻﺔ ﻣﻦ ﻧﻤﺎذج ‪ ARIMA‬ﻓﺈن ﻃﺮق‬
‫اﻟﺘﻌﺎﻣﻞ ﻣﻌﻬﺎ هﻲ ﻧﻔﺲ اﻟﻄﺮق اﻟﺴﺎﺑﻘﺔ ﻣﻦ ﺣﻴﺚ اﻟﺘﻌﺮف ﻋﻠﻲ ﺷﻜﻞ اﻟﻨﻤﻮذج وﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ اﻟﻨﻤﻮذج‬
‫واﻹﺧﺘﺒﺎرات اﻟﺘﻔﺤﺼﻴﺔ وﻣﻦ ﺛﻢ اﻟﺘﻨﺒﺆ‪ .‬ﺟﻤﻴﻊ اﻟﻄﺮق واﻟﻤﻌﺎدﻻت اﻟﺘﻲ درﺳﻨﺎهﺎ ﺳﺎﺑﻘﺎ ﻟﻠﻨﻤﺎذج ﻏﻴﺮ‬
‫اﻟﻤﻮﺳﻤﻴﺔ ﺗﻨﻄﺒﻖ هﻨﺎ‪ .‬ﺳﻮف ﻧﺸﺘﻖ دوال اﻟﺘﻨﺒﺆ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ﻟﻠﺘﻮﺿﻴﺢ ﻓﻘﻂ‪.‬‬
‫‪ -١‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻠﻨﻤﻮذج ‪: SARIMA(٠،٠،٠)(٠،١،١)١٢‬‬
‫وﻳﻜﺘﺐ اﻟﻨﻤﻮذج ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫‪(1 − B ) z = (1 − ΘB ) a‬‬
‫‪12‬‬

‫‪t‬‬

‫‪12‬‬

‫‪t‬‬

‫ﻣﻦ اﻟﻤﻌﺎدﻟﺔ اﻟﻔﺮوﻗﻴﺔ‬
‫‪zn + l = zn +l−12 + an + l − Θan + l−12‬‬

‫ﻳﻤﻜﻦ اﻟﺤﺼﻮل ﻋﻠﻰ اﻟﺘﻨﺒﺆات آﺎﻟﺘﺎﻟﻲ‬

‫‪zn (1) = zn −11 − Θan −11‬‬
‫‪zn ( 2 ) = zn −10 − Θan −10‬‬

‫‪M‬‬

‫‪zn (12 ) = zn − Θan‬‬
‫‪zn ( l ) = zn ( l − 12 ) , l ≥ 12‬‬

‫أو‬
‫‪l = 1, 2,...,12‬‬

‫‪l > 12‬‬

‫واﺿﺢ أن‬

‫‪⎧ zn + l−12 − Θan + l−12 ,‬‬
‫⎨ = ) ‪zn ( l‬‬
‫‪⎩ zn ( l − 12 ) ,‬‬
‫‪zn (1) = zn (13) = zn ( 25) = L‬‬

‫‪zn ( 2 ) = zn (14 ) = zn ( 26 ) = L‬‬

‫‪M‬‬

‫‪zn (12 ) = zn ( 24 ) = zn ( 36 ) = L‬‬

‫ﺗﺒﺎﻳﻦ أﺧﻄﺎء اﻟﺘﻨﺒﺆ‬

‫)‬

‫‪2‬‬
‫‪l −1‬‬

‫‪+ L +ψ‬‬

‫‪2‬‬
‫‪1‬‬

‫‪(1 + ψ‬‬

‫‪2‬‬

‫‪V ⎣⎡ en ( l )⎦⎤ = σ‬‬

‫وداﻟﺔ اﻷوزان ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ )ﺑﺮهﻦ ذﻟﻚ(‬
‫‪j = 12, 24,...‬‬
‫‪otherwise‬‬

‫‪⎧1 − Θ,‬‬
‫‪⎩ 0,‬‬

‫⎨= ‪ψj‬‬

‫وﺑﺘﻌﻮﻳﺾ اﻷوزان ﻓﻲ ﺻﻴﻐﺔ ﺗﺒﺎﻳﻦ أﺧﻄﺎء اﻟﺘﻨﺒﺆ ﻧﺠﺪ‬
‫⎥‪⎧ ⎢ l − 1‬‬
‫⎫‪2‬‬
‫⎢ ‪V ⎡⎣ en ( l )⎤⎦ = σ 2 ⎨1 +‬‬
‫⎬ ) ‪(1 − Θ‬‬
‫⎥‬
‫⎦ ‪⎩ ⎣ 12‬‬
‫⎭‬

‫ﺣﻴﺚ ⎦⎥ ‪ ⎢⎣ x‬ﺗﻌﻨﻲ اﻟﺠﺰء اﻟﺼﺤﻴﺢ ﻣﻦ ‪. x‬‬
‫‪ -٢‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻠﻨﻤﻮذج ‪: SARIMA(٠،١،١)(٠،١،١)١٢‬‬
‫وﻳﻜﺘﺐ اﻟﻨﻤﻮذج ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫‪(1 − B ) (1 − B12 ) zt = (1 − θ B ) (1 − ΘB12 ) at‬‬

‫‪١١٧‬‬

‫ﻣﻦ اﻟﻤﻌﺎدﻟﺔ اﻟﻔﺮوﻗﻴﺔ‬

‫‪zn + l = zn +l−1 + zn + l−12 − zn + l−13 + an + l − θ an +l−1 − Θan + l−12 + θ Θan+ l−13‬‬

‫ﻳﻤﻜﻦ اﻟﺤﺼﻮل ﻋﻠﻰ اﻟﺘﻨﺒﺆات آﺎﻟﺘﺎﻟﻲ‬

‫‪zn (1) = zn + zn −11 − zn −13 − θ an − Θan −11 + θ Θan −12‬‬
‫‪zn ( 2 ) = zn (1) + zn −10 − zn −11 − Θan −10 + θ Θan −11‬‬
‫‪M‬‬

‫‪zn (12 ) = zn (11) + zn − zn −1 − Θan + θ Θan −1‬‬
‫‪zn (13) = zn (12 ) + zn (1) − zn + θ Θan‬‬
‫)‪zn ( l ) = zn ( l − 1) + zn ( l − 12 ) − zn ( l − 13‬‬

‫وهﻜﺬا ﺑﻘﻴﻢ أوﻟﻴﺔ‬

‫‪zn (1) = zn + zn −11 − zn −13 − θ an − Θan −11 + θ Θan −12‬‬
‫‪zn ( 2 ) = zn (1) + zn −10 − zn −11 − Θan −10 + θ Θan −11‬‬

‫‪M‬‬

‫‪zn (12 ) = zn (11) + zn − zn −1 − Θan + θ Θan −1‬‬
‫‪zn (13) = zn (12 ) + zn (1) − zn + θ Θan‬‬

‫وﻋﻼﻗﺔ ﺗﻜﺮارﻳﺔ‬

‫‪zn ( l ) = zn ( l − 1) + zn ( l − 12 ) − zn ( l − 13) , l > 13‬‬

‫ﻳﻤﻜﻦ ﺗﻮﻟﻴﺪ اﻟﻌﺪد اﻟﻤﻄﻠﻮب ﻣﻦ اﻟﺘﻨﺒﺆات‪.‬‬
‫أﻣﺜﻠﺔ وﺣﺎﻻت دراﺳﺔ ﻟﺒﻌﺾ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ‪:‬‬
‫ﻣﺜﺎل )‪ : (١‬ﺳﻮف ﻧﺤﺎول إﻳﺠﺎد ﻧﻤﻮذج ﻣﻦ ﻋﺎﺋﻠﺔ ‪ SARIMA‬ﻳﻨﻄﺒﻖ ﻋﻠﻰ اﻟﻤﺸﺎهﺪات اﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪653‬‬
‫‪736‬‬
‫‪615‬‬
‫‪688‬‬
‫‪775‬‬
‫‪845‬‬
‫‪740‬‬
‫‪800‬‬
‫‪883‬‬
‫‪947‬‬
‫‪817‬‬

‫‪566‬‬
‫‪770‬‬
‫‪653‬‬
‫‪645‬‬
‫‪696‬‬
‫‪871‬‬
‫‪783‬‬
‫‪763‬‬
‫‪773‬‬
‫‪969‬‬
‫‪858‬‬

‫‪600‬‬
‫‪705‬‬
‫‪702‬‬
‫‪667‬‬
‫‪717‬‬
‫‪805‬‬
‫‪819‬‬
‫‪800‬‬
‫‪821‬‬
‫‪902‬‬
‫‪896‬‬

‫‪582‬‬
‫‪688‬‬
‫‪756‬‬
‫‪661‬‬
‫‪698‬‬
‫‪785‬‬
‫‪859‬‬
‫‪790‬‬
‫‪805‬‬
‫‪889‬‬
‫‪937‬‬

‫‪553‬‬
‫‪618‬‬
‫‪782‬‬
‫‪697‬‬
‫‪660‬‬
‫‪690‬‬
‫‪886‬‬
‫‪834‬‬
‫‪766‬‬
‫‪778‬‬
‫‪966‬‬

‫‪577‬‬
‫‪628‬‬
‫‪722‬‬
‫‪735‬‬
‫‪687‬‬
‫‪734‬‬
‫‪824‬‬
‫‪869‬‬
‫‪810‬‬
‫‪828‬‬
‫‪903‬‬

‫‪568‬‬
‫‪598‬‬
‫‪709‬‬
‫‪798‬‬
‫‪681‬‬
‫‪711‬‬
‫‪807‬‬
‫‪913‬‬
‫‪809‬‬
‫‪802‬‬
‫‪892‬‬

‫‪599‬‬
‫‪565‬‬
‫‪622‬‬
‫‪811‬‬
‫‪722‬‬
‫‪677‬‬
‫‪707‬‬
‫‪942‬‬
‫‪855‬‬
‫‪760‬‬
‫‪782‬‬

‫واﻟﻤﺨﻄﻂ اﻟﺰﻣﻨﻲ ﻟﻠﻤﺸﺎهﺪات‬

‫‪١١٨‬‬

‫‪640‬‬
‫‪587‬‬
‫‪658‬‬
‫‪755‬‬
‫‪767‬‬
‫‪706‬‬
‫‪750‬‬
‫‪878‬‬
‫‪894‬‬
‫‪791‬‬
‫‪834‬‬

‫‪697‬‬
‫‪583‬‬
‫‪634‬‬
‫‪736‬‬
‫‪817‬‬
‫‪701‬‬
‫‪734‬‬
‫‪860‬‬
‫‪935‬‬
‫‪784‬‬
‫‪813‬‬

‫‪727‬‬
‫‪617‬‬
‫‪594‬‬
‫‪635‬‬
‫‪837‬‬
‫‪740‬‬
‫‪690‬‬
‫‪756‬‬
‫‪961‬‬
‫‪837‬‬
‫‪773‬‬

‫‪656‬‬
‫‪660‬‬
‫‪611‬‬
‫‪677‬‬
‫‪784‬‬
‫‪783‬‬
‫‪723‬‬
‫‪804‬‬
‫‪900‬‬
‫‪881‬‬
‫‪812‬‬

‫‪640‬‬
‫‪716‬‬
‫‪604‬‬
‫‪635‬‬
‫‪762‬‬
‫‪826‬‬
‫‪725‬‬
‫‪751‬‬
‫‪890‬‬
‫‪924‬‬
‫‪815‬‬
‫‪843‬‬

‫‪561‬‬
‫‪742‬‬
‫‪639‬‬
‫‪602‬‬
‫‪667‬‬
‫‪858‬‬
‫‪764‬‬
‫‪711‬‬
‫‪799‬‬
‫‪957‬‬
‫‪867‬‬
‫‪797‬‬

‫)‪z(t‬‬
‫‪589‬‬
‫‪673‬‬
‫‪678‬‬
‫‪621‬‬
‫‪713‬‬
‫‪796‬‬
‫‪801‬‬
‫‪747‬‬
‫‪826‬‬
‫‪898‬‬
‫‪908‬‬
‫‪827‬‬

‫‪1000‬‬

‫‪900‬‬

‫)‪z(t‬‬

‫‪800‬‬

‫‪700‬‬

‫‪600‬‬

‫‪150‬‬

‫‪50‬‬

‫‪100‬‬

‫‪In d e x‬‬

‫ﻳﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﺘﺒﺎﻳﻦ واﻟﻤﺘﻮﺳﻂ ﻟﺬﻟﻚ ﻧﺜﺒﺖ اﻟﺘﺒﺎﻳﻦ أوﻻ ﺑﺘﺤﻮﻳﻞ ﻟﻮﻏﺎرﺛﻤﻲ‬
‫أي ) ‪ yt = ln ( zt‬وﻧﺮﺳﻢ اﻟﻤﺨﻄﻂ اﻟﺰﻣﻨﻲ ﻟﻬﺎ‬
‫‪6 .9‬‬
‫‪6 .8‬‬
‫‪6 .7‬‬

‫)‪y(t‬‬

‫‪6 .6‬‬
‫‪6 .5‬‬
‫‪6 .4‬‬
‫‪6 .3‬‬

‫‪1 5 0‬‬

‫‪5 0‬‬

‫‪1 0 0‬‬

‫‪In d e x‬‬

‫ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ اﺳﺘﻘﺮت ﻓﻲ اﻟﺘﺒﺎﻳﻦ وﻟﻜﻦ ﻻﺗﺰال ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ ﻟﺬﻟﻚ ﻧﺄﺧﺬ اﻟﻔﺮق‬
‫اﻷول ) ‪ xt = (1 − B ) yt = (1 − B ) ln ( zt‬وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬
‫‪0 .1 5‬‬
‫‪0 .1 0‬‬

‫‪0 .0 0‬‬

‫)‪y(t)-y(t-1‬‬

‫‪0 .0 5‬‬

‫‪-0 .0 5‬‬
‫‪-0 .1 0‬‬
‫‪1 5 0‬‬

‫‪1 0 0‬‬

‫‪5 0‬‬

‫‪In d e x‬‬

‫اﻟﻤﺘﺴﻠﺴﻠﺔ اﻵن ﻣﺴﺘﻘﺮة ﻓﻲ آﻞ ﻣﻦ اﻟﺘﺒﺎﻳﻦ واﻟﻤﺘﻮﺳﻂ‪ .‬ﻟﻨﻨﻈﺮ إﻟﻰ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ‬
‫اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻬﺎ‬

‫‪١١٩‬‬

Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r y ( t ) - y ( t
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
- 0 .2
- 0 .4
- 0 .6
- 0 .8
- 1 .0

10

20

30

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

1
2
3
4
5
6
7
8
9
10
11
12

0 .0 1
0 .2 5
- 0 .0 8
- 0 .3 7
- 0 .0 6
- 0 .5 0
- 0 .0 4
- 0 .3 5
- 0 .0 5
0 .2 3
0 .0 1
0 .9 0

0 .1 2
3 .2 3
- 1 .0 1
- 4 .4 8
- 0 .6 8
- 5 .4 1
- 0 .3 3
- 3 .3 0
- 0 .4 3
2 .0 1
0 .1 1
7 .7 4

0 .0 2
1 0 .7 0
1 1 .8 8
3 5 .6 0
3 6 .2 8
7 9 .9 7
8 0 .1 9
1 0 2 .3 7
1 0 2 .8 0
1 1 2 .2 9
1 1 2 .3 2
2 6 1 .4 1

13
14
15
16
17
18
19
20
21
22
23
24

0 .0 2
0 .2 3
- 0 .0 7
- 0 .3 4
- 0 .0 6
- 0 .4 6
- 0 .0 3
- 0 .3 2
- 0 .0 5
0 .2 1
0 .0 1
0 .8 2

0 .1 2
1 .5 1
- 0 .4 7
- 2 .1 8
- 0 .3 9
- 2 .8 8
- 0 .2 0
- 1 .9 3
- 0 .2 6
1 .2 0
0 .0 8
4 .7 6

2 6 1 .4 7
2 7 1 .2 9
2 7 2 .3 0
2 9 3 .7 3
2 9 4 .4 5
3 3 4 .6 7
3 3 4 .8 8
3 5 4 .9 6
3 5 5 .3 6
3 6 3 .5 5
3 6 3 .5 8
4 9 7 .3 0

25
26
27
28
29
30
31
32
33
34
35
36

0 .0 2
0 .2 1
- 0 .0 6
- 0 .3 1
- 0 .0 6
- 0 .4 2
- 0 .0 2
- 0 .3 0
- 0 .0 5
0 .1 8
0 .0 0
0 .7 6

0 .0 8
1 .1 0
- 0 .2 9
- 1 .5 7
- 0 .2 9
- 2 .0 9
- 0 .1 2
- 1 .4 7
- 0 .2 4
0 .8 9
0 .0 2
3 .6 3

4 9 7 .3 5
5 0 6 .5 0
5 0 7 .1 6
5 2 6 .4 5
5 2 7 .1 4
5 6 3 .0 5
5 6 3 .1 6
5 8 2 .0 5
5 8 2 .5 8
5 8 9 .8 4
5 8 9 .8 5
7 1 3 .1 4

40

L ag

T

LBQ

3 7 0 .0 1 0 .0 5
3 8 0 .2 0 0 .9 0
3 9 - 0 .0 4 - 0 .1 9
4 0 - 0 .2 7 - 1 .2 1
4 1 - 0 .0 5 - 0 .2 4

C o rr

7 1 3 .1 6
7 2 2 .0 6
7 2 2 .4 5
7 3 8 .8 6
7 3 9 .5 1

Partial Autocorrelation

P a r t ia l A u t o c o r r e la t i o n F u n c t io n f o r y ( t ) - y ( t
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

10

20

30

40

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

1
2
3
4
5
6
7
8
9
10
11
12

0 .0 1
0 .2 5
- 0 .0 9
- 0 .4 6
- 0 .0 2
- 0 .3 5
- 0 .1 4
- 0 .4 8
- 0 .4 1
- 0 .2 3
- 0 .5 8
0 .6 3

0 .1 2
3 .2 3
- 1 .2 0
- 5 .9 7
- 0 .2 2
- 4 .5 2
- 1 .8 2
- 6 .2 6
- 5 .3 6
- 2 .9 9
- 7 .5 5
8 .0 8

13
14
15
16
17
18
19
20
21
22
23
24

-0 .0 4
-0 .3 3
0 .0 0
0 .1 8
0 .0 1
0 .0 8
-0 .0 8
0 .0 8
0 .0 2
-0 .0 4
0 .0 2
0 .0 8

- 0 .4 8
- 4 .2 6
0 .0 6
2 .2 9
0 .0 7
1 .0 6
- 1 .0 4
0 .9 8
0 .2 9
- 0 .5 3
0 .2 9
1 .0 4

25
26
27
28
29
30
31
32
33
34
35
36

- 0 .1 3
- 0 .0 5
0 .0 6
0 .0 0
- 0 .0 7
0 .0 3
0 .0 7
- 0 .0 1
- 0 .1 0
0 .0 6
0 .0 3
0 .0 1

-1 .6 7
-0 .6 1
0 .8 4
0 .0 5
-0 .9 4
0 .3 4
0 .9 3
-0 .0 9
-1 .3 1
0 .8 0
0 .3 9
0 .1 3

37
38
39
40
41

- 0 .0 4
0 .0 2
0 .0 4
0 .0 6
- 0 .0 2

- 0 .5 1
0 .3 0
0 .4 8
0 .8 1
- 0 .2 0

‫ و‬١٢ ‫ﻳﻼﺣﻆ ﻣﻦ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ان اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻣﻮﺳﻤﻴﺎ ﻷن ﻗﻴﻤﻬﺎ ﻋﻨﺪ اﻟﺘﺨﻠﻔﺎت‬
wt = (1 − B12 ) (1 − B ) ln ( zt ) ‫ ﺗﺘﺨﺎﻣﺪ ﺑﺒﻂء ﻟﺬﻟﻚ ﻧﺄﺧﺬ اﻟﻔﺮق اﻟﻤﻮﺳﻤﻲ اﻷول‬٣٦ ‫ و‬٢٤
‫وﻧﺮﺳﻤﻬﺎ ﺑﻌﺪ هﺬا اﻟﺘﻔﺮﻳﻖ‬

y(t)-y(t-1)12

0 .0 5

0 .0 0

-0 .0 5
In d e x

50

100

150

‫ﻧﻮﺟﺪ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻬﺎ‬

١٢٠

Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r y ( t ) - y ( t
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
- 0 .2
- 0 .4
- 0 .6
- 0 .8
- 1 .0

5

15

25

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

1
2
3
4
5
6
7
8
9
10
11
12

- 0 .2 1
- 0 .0 1
0 .1 0
- 0 .1 3
- 0 .1 0
- 0 .0 2
0 .1 2
0 .0 5
- 0 .0 5
0 .1 3
- 0 .0 1
- 0 .4 4

- 2 .6 5
- 0 .1 3
1 .1 4
- 1 .4 9
- 1 .1 1
- 0 .2 8
1 .3 3
0 .5 4
- 0 .5 8
1 .4 5
- 0 .0 9
- 4 .9 0

7 .1 5
7 .1 7
8 .6 3
1 1 .2 0
1 2 .6 7
1 2 .7 7
1 4 .9 5
1 5 .3 1
1 5 .7 5
1 8 .4 9
1 8 .5 0
5 0 .8 8

13
14
15
16
17
18
19
20
21
22
23
24

0 .1 8
- 0 .0 7
- 0 .0 5
0 .0 3
0 .1 2
- 0 .0 0
- 0 .1 1
0 .0 3
- 0 .0 2
- 0 .0 9
0 .1 1
- 0 .0 4

1 .8 0
- 0 .7 0
- 0 .4 8
0 .2 6
1 .1 3
- 0 .0 2
- 1 .0 8
0 .2 5
- 0 .2 0
- 0 .8 0
0 .9 9
- 0 .3 8

5 6 .6 4
5 7 .5 4
5 7 .9 7
5 8 .1 0
6 0 .5 8
6 0 .5 8
6 2 .9 1
6 3 .0 4
6 3 .1 2
6 4 .4 5
6 6 .5 2
6 6 .8 2

25
26
27
28
29
30
31
32
33
34
35
36

0 .0 7
- 0 .0 0
- 0 .0 6
0 .0 3
- 0 .1 0
0 .0 1
0 .0 4
0 .0 0
0 .0 2
0 .0 0
0 .0 9
- 0 .0 6

0 .6 2
- 0 .0 2
- 0 .5 7
0 .3 0
- 0 .9 0
0 .0 8
0 .3 8
0 .0 4
0 .1 9
0 .0 3
0 .8 0
- 0 .5 1

6 7 .6 7
6 7 .6 7
6 8 .3 8
6 8 .5 8
7 0 .4 0
7 0 .4 1
7 0 .7 4
7 0 .7 4
7 0 .8 3
7 0 .8 3
7 2 .3 6
7 2 .9 9

35
L ag

T

LBQ

3 7 - 0 .0 6 - 0 .5 4
3 8 0 .0 3 0 .2 6

C o rr

7 3 .7 1
7 3 .8 8

Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r y ( t ) - y ( t
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

5

15

25

35

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

1
2
3
4
5
6
7
8
9
10
11
12

- 0 .2 1
- 0 .0 6
0 .0 8
- 0 .0 9
- 0 .1 5
- 0 .1 0
0 .1 1
0 .1 1
- 0 .0 4
0 .0 6
0 .0 5
- 0 .4 2

- 2 .6 5
- 0 .7 3
1 .0 6
- 1 .1 6
- 1 .8 2
- 1 .1 9
1 .4 1
1 .4 2
- 0 .5 1
0 .8 0
0 .5 9
- 5 .2 3

13
14
15
16
17
18
19
20
21
22
23
24

-0 .0 0
-0 .0 2
0 .0 0
-0 .1 1
0 .0 4
0 .0 1
-0 .0 4
-0 .0 3
-0 .0 3
0 .0 3
0 .0 7
-0 .2 7

- 0 .0 3
- 0 .2 5
0 .0 0
- 1 .4 0
0 .4 7
0 .1 2
- 0 .5 2
- 0 .3 5
- 0 .3 1
0 .3 2
0 .9 1
- 3 .4 1

25
26
27
28
29
30
31
32
33
34
35
36

0 .1 2
- 0 .0 1
- 0 .1 2
- 0 .0 7
0 .0 4
- 0 .0 4
- 0 .0 5
0 .0 2
- 0 .0 1
- 0 .0 2
0 .1 8
- 0 .2 1

1 .5 1
-0 .1 5
-1 .4 4
-0 .9 1
0 .4 4
-0 .5 3
-0 .6 0
0 .1 9
-0 .1 2
-0 .1 9
2 .2 1
-2 .5 7

37
38

0 .1 0
- 0 .0 6

1 .2 3
- 0 .7 9

wt = (1 − B12 ) (1 − B ) ln ( zt ) ‫ﻣﻦ اﻧﻤﺎط دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬

‫ اي ﻧﻄﺒﻖ اﻟﻨﻤﻮذج‬٠ ‫ و‬٠ ‫ اﻟﻤﻤﻜﻨﺔ هﻲ‬q ‫ و‬p ‫ﻧﺠﺪ ان ﻗﻴﻢ‬

(1 − B ) (1 − B ) ln ( z ) = (1 − ΘB ) a
12

12

t

t

‫ ﻳﻄﺒﻖ هﺬا اﻟﻨﻤﻮذج‬MINITAB ‫ اﻷﻣﺮ اﻟﺘﺎﻟﻲ ﻓﻲ‬SARIMA(٠،١،٠)(٠،١،١)١٢ ‫هﻮ‬
ARIMA 0 1 ٠ 0 1 1 12 'y(t)' ;
NoConstant.
zt = e y ‫ وﻟﻠﺤﺼﻮل ﻋﻠﻰ اﻟﻨﺘﺎﺋﺞ اﻟﻨﻬﺎﺋﻴﺔ ﻧﺠﺮي اﻟﺘﺤﻮﻳﻞ‬yt = ln ( zt ) ‫ﻻﺣﻆ اﻧﻨﺎ اﺳﺘﺨﺪﻣﻨﺎ‬
:‫اﻟﻨﺘﺎﺋﺞ‬
t

MTB > Name c14 = 'RESI3' c15 = 'FITS3'
MTB > ARIMA 0 1 0 0 1 1 12 'y(t)' 'RESI3' 'FITS3';
SUBC>
NoConstant;
SUBC>
Forecast 24 c7 c8 c9;
SUBC> GACF;
SUBC> GPACF;
SUBC> GHistogram;
SUBC> GNormalplot.

ARIMA Model

١٢١

ARIMA model for y(t)
Estimates at each iteration
Iteration
SSE
Parameters
0
0.0228597
0.100
1
0.0204943
0.250
2
0.0187066
0.400
3
0.0174234
0.550
4
0.0169841
0.684
5
0.0169841
0.683
6
0.0169841
0.683
Relative change in each estimate less than
Final Estimates of Parameters
Type
Coef
StDev
SMA 12
0.6831
0.0610

0.0010

T
11.20

Differencing: 1 regular, 1 seasonal of order 12
Number of observations: Original series 168, after differencing 155
Residuals:
SS = 0.0165799 (backforecasts excluded)
MS = 0.0001077 DF = 154
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
9.0(DF=11)
29.9(DF=23)
44.5(DF=35)

48
59.4(DF=47)

Forecasts from period 168
Period
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192

Forecast
6.76750
6.70901
6.83815
6.85381
6.92288
6.89349
6.84654
6.80008
6.74395
6.75028
6.70664
6.75999
6.79052
6.73203
6.86117
6.87684
6.94590
6.91651
6.86956
6.82310
6.76697
6.77330
6.72966
6.78301

95 Percent Limits
Lower
Upper
6.74716
6.78784
6.68024
6.73778
6.80292
6.87338
6.81313
6.89450
6.87739
6.96836
6.84366
6.94331
6.79272
6.90035
6.74255
6.85761
6.68293
6.80497
6.68596
6.81461
6.63918
6.77410
6.68952
6.83045
6.71514
6.86590
6.65203
6.81203
6.77680
6.94554
6.78832
6.96535
6.85342
7.03838
6.82023
7.01279
6.76962
6.96950
6.71963
6.92657
6.66009
6.87385
6.66312
6.88349
6.61627
6.84305
6.66649
6.89952

(1 − B ) (1 − B ) ln ( z ) = (1 − 0.683B ) a ,
12

12

t

t

Actual

‫أي أن اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻟﻬﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ هﻮ‬

at

N ( 0, 0.0001077 )

‫ﻻﺣﻆ ان‬

Θ = 0.683, s.e. ( Θ ) = 0.061, t = 11.2

.‫أي ان اﻟﻤﻌﻠﻢ ﻋﺎﻟﻲ اﻟﻤﻌﻨﻮﻳﺔ‬
:‫ﻓﺤﺺ اﻟﺒﻮاﻗﻲ‬
‫إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ‬
MTB > ZTest 0.0 0.0103778 'RESI3';
SUBC>
Alternative 0.

١٢٢

Z-Test
Test of mu = 0.000000 vs mu not = 0.000000
The assumed sigma = 0.0104
Variable
RESI3

N
Mean
StDev
155 -0.000111 0.010375

SE Mean
0.000834

Z
-0.13

P
0.89

‫ اي ﻻﻧﺮﻓﺾ أن ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ ﺻﻔﺮا‬٠٫٠٥ ‫ وهﻲ اآﺒﺮ ﻣﻦ‬P-value=٠٫٨٩ ‫ﻻﺣﻆ ان اﻟـ‬
‫إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬
MTB > Runs 0 'RESI3'.

Runs Test
RESI3
K =

0.0000

The observed number of runs = 70
The expected number of runs = 78.1097
72 Observations above K
83 below
The test is significant at 0.1893
Cannot reject at alpha = 0.05

‫ اي اﻧﻨﺎ ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬٠٫١٨٩٣ ‫اﻹﺧﺘﺒﺎر ﻣﻌﻨﻮي ﻋﻨﺪ‬
:‫إﺧﺘﺒﺎر إﺳﺘﻘﻼل اﻟﺒﻮاﻗﻲ‬
‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬
A C F o f R e s id u a ls f o r y ( t )
( w it h 9 5 % c o n f id e n c e l im it s f o r t h e a u t o c o r r e l a t io n s )
1 .0
0 .8

Autocorrelation

0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0
3

6

9

12

15

18

21

Lag

١٢٣

24

27

30

33

36

39

P A C F o f R e s id u a ls f o r y ( t )
( w it h 9 5 % c o n f id e n c e l im it s f o r t h e p a r t ia l a u t o c o r r e l a t io n s )
1 .0

0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0
3

6

9

12

15

18

21

24

27

30

33

36

39

Lag

.‫ﻧﻼﺣﻆ اﻧﻬﺎ ﺗﻌﻄﻲ اﻧﻤﺎط اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء أي اﻧﻬﺎ ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ وإذا آﺎﻧﺖ ﻃﺒﻴﻌﻴﺔ ﻓﻬﻲ ﻣﺴﺘﻘﻠﺔ‬
:‫إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬
Histogram of the Residuals
(response is y(t))

Frequency

30

20

10

0
-0.03

-0.02

-0.01

0.00

0.01

0.02

0.03

0.04

Residual

Normal Probability Plot of the Residuals
(response is y(t))
0.04
0.03
0.02

Residual

Partial Autocorrelation

0 .8

0.01
0.00
-0.01
-0.02
-0.03
-3

-2

-1

0

1

2

Normal Score

١٢٤

3

‫‪K-S test for Residuals‬‬

‫‪.999‬‬
‫‪.99‬‬
‫‪.95‬‬

‫‪.50‬‬
‫‪.20‬‬

‫‪Probability‬‬

‫‪.80‬‬

‫‪.05‬‬
‫‪.01‬‬
‫‪.001‬‬
‫‪0.03‬‬

‫‪0.02‬‬

‫‪0.01‬‬

‫‪-0.01‬‬

‫‪0.00‬‬

‫‪-0.02‬‬

‫‪-0.03‬‬

‫‪RESI3‬‬
‫‪Average: -0.0001115‬‬
‫‪StDev: 0.0103754‬‬
‫‪N: 155‬‬

‫‪Kolmogorov-Smirnov Normality Test‬‬
‫‪D+: 0.074 D-: 0.045 D : 0.074‬‬
‫‪Approximate P-Value: 0.041‬‬

‫ﻻﺣﻆ ان اﻟـ ‪ P-value‬ﻹﺧﺘﺒﺎر ‪ K-S‬ﻳﻌﻄﻲ ‪ ٠٫٠٤١‬وهﻲ اﻗﻞ ﻣﻦ ‪ ٠٫٠٥‬اذا اﻹﺧﺘﺒﺎر ﻣﻌﻨﻮي ﻋﻨﺪ‬
‫‪ α = 0.05‬اي ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‪.‬‬
‫اﻟﺘﻨﺒﺆ ﺑﺈﺳﺘﺨﺪام اﻟﻨﻤﻮذج‪:‬‬
‫ﻓﻲ اﻟﻤﺨﺮﺟﺎت اﻟﺴﺎﺑﻘﺔ ﻗﻤﻨﺎ ﺑﺎﻟﺘﻨﺒﺆ ﻋﻦ ‪ ٢٤‬ﻗﻴﻤﺔ ﻣﺴﺘﻘﺒﻠﻴﺔ ﻣﻊ ‪ ٩٥%‬ﻓﺘﺮات ﺗﻨﺒﺆ وﻧﺮﺳﻤﻬﺎ ﺑﺎﻟﺮﺳﻢ‬
‫اﻟﺘﺎﻟﻲ‪:‬‬
‫‪1150‬‬

‫‪1050‬‬

‫‪850‬‬

‫‪750‬‬

‫‪25‬‬

‫‪20‬‬

‫‪10‬‬

‫‪15‬‬

‫‪T im e‬‬

‫واﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﻣﻊ اﻟﺘﻨﺒﺆات وﺣﺪود اﻟﺘﻨﺒﺆ‬

‫‪١٢٥‬‬

‫‪5‬‬

‫‪0‬‬

‫‪Forecast‬‬

‫‪950‬‬

‫‪1150‬‬

‫‪1050‬‬

‫‪850‬‬
‫‪750‬‬

‫‪Forecast‬‬

‫‪950‬‬

‫‪650‬‬
‫‪550‬‬
‫‪200‬‬

‫‪0‬‬

‫‪100‬‬

‫‪Time‬‬

‫ﺣﺎﻟﺔ دراﺳﺔ ‪: ١‬‬
‫ﺳﻮف ﻧﺤﺎول إﻳﺠﺎد ﻧﻤﻮذج ﻣﻦ ﻋﺎﺋﻠﺔ ‪ SARIMA‬ﻳﻨﻄﺒﻖ ﻋﻠﻰ اﻟﻤﺸﺎهﺪات اﻟﺘﺎﻟﻴﺔ‪:‬‬
‫‪57.4‬‬
‫‪60.0‬‬
‫‪69.8‬‬
‫‪70.1‬‬
‫‪66.1‬‬
‫‪58.7‬‬
‫‪55.8‬‬
‫‪51.6‬‬
‫‪53.5‬‬
‫‪53.0‬‬
‫‪52.1‬‬
‫‪53.6‬‬
‫‪55.6‬‬
‫‪62.6‬‬
‫‪75.5‬‬
‫‪76.4‬‬

‫‪61.5‬‬
‫‪71.0‬‬
‫‪71.0‬‬
‫‪67.5‬‬
‫‪56.1‬‬
‫‪55.4‬‬
‫‪55.3‬‬
‫‪52.4‬‬
‫‪53.4‬‬
‫‪53.5‬‬
‫‪53.4‬‬
‫‪55.3‬‬
‫‪60.1‬‬
‫‪70.3‬‬
‫‪73.3‬‬
‫‪70.8‬‬

‫‪72.7‬‬
‫‪70.6‬‬
‫‪67.4‬‬
‫‪58.0‬‬
‫‪53.6‬‬
‫‪52.9‬‬
‫‪53.2‬‬
‫‪52.1‬‬
‫‪53.6‬‬
‫‪54.3‬‬
‫‪55.3‬‬
‫‪58.5‬‬
‫‪68.9‬‬
‫‪69.4‬‬
‫‪68.1‬‬
‫‪60.6‬‬

‫‪72.2‬‬
‫‪68.2‬‬
‫‪58.2‬‬
‫‪54.2‬‬
‫‪53.0‬‬
‫‪53.0‬‬
‫‪52.8‬‬
‫‪52.6‬‬
‫‪55.1‬‬
‫‪56.3‬‬
‫‪58.2‬‬
‫‪69.3‬‬
‫‪68.8‬‬
‫‪64.7‬‬
‫‪58.1‬‬
‫‪56.4‬‬

‫‪71.5‬‬
‫‪57.7‬‬
‫‪54.3‬‬
‫‪53.2‬‬
‫‪52.8‬‬
‫‪52.7‬‬
‫‪52.8‬‬
‫‪53.9‬‬
‫‪57.3‬‬
‫‪59.4‬‬
‫‪66.9‬‬
‫‪69.6‬‬
‫‪63.6‬‬
‫‪57.1‬‬
‫‪54.8‬‬
‫‪55.6‬‬

‫‪59.1‬‬
‫‪54.6‬‬
‫‪53.4‬‬
‫‪53.0‬‬
‫‪52.6‬‬
‫‪53.4‬‬
‫‪54.5‬‬
‫‪56.4‬‬
‫‪61.9‬‬
‫‪70.4‬‬
‫‪70.7‬‬
‫‪64.2‬‬
‫‪57.1‬‬
‫‪53.9‬‬
‫‪54.3‬‬
‫‪55.0‬‬

‫‪57.2‬‬
‫‪54.9‬‬
‫‪53.0‬‬
‫‪53.0‬‬
‫‪52.9‬‬
‫‪54.9‬‬
‫‪56.4‬‬
‫‪61.7‬‬
‫‪69.9‬‬
‫‪69.9‬‬
‫‪65.3‬‬
‫‪55.5‬‬
‫‪52.2‬‬
‫‪53.5‬‬
‫‪54.6‬‬
‫‪54.7‬‬

‫‪56.3‬‬
‫‪54.9‬‬
‫‪52.8‬‬
‫‪53.4‬‬
‫‪54.0‬‬
‫‪57.5‬‬
‫‪59.3‬‬
‫‪68.3‬‬
‫‪71.5‬‬
‫‪61.0‬‬
‫‪56.5‬‬
‫‪53.3‬‬
‫‪51.5‬‬
‫‪53.5‬‬
‫‪54.2‬‬
‫‪55.8‬‬

‫‪55.7‬‬
‫‪55.3‬‬
‫‪54.4‬‬
‫‪55.6‬‬
‫‪56.6‬‬
‫‪69.3‬‬
‫‪70.0‬‬
‫‪65.3‬‬
‫‪57.3‬‬
‫‪53.9‬‬
‫‪52.5‬‬
‫‪51.5‬‬
‫‪51.5‬‬
‫‪53.3‬‬
‫‪55.8‬‬
‫‪60.5‬‬
‫‪62.2‬‬

‫‪55.8‬‬
‫‪54.9‬‬
‫‪53.3‬‬
‫‪54.6‬‬
‫‪54.9‬‬
‫‪61.2‬‬
‫‪68.7‬‬
‫‪67.9‬‬
‫‪65.1‬‬
‫‪52.7‬‬
‫‪53.4‬‬
‫‪52.4‬‬
‫‪51.7‬‬
‫‪53.1‬‬
‫‪54.8‬‬
‫‪57.7‬‬

‫اﻟﻤﺨﻄﻂ اﻟﺰﻣﻨﻲ‬

‫‪7 0‬‬

‫)‪z(t‬‬
‫‪6 0‬‬

‫‪5 0‬‬
‫‪1 5 0‬‬

‫‪5 0‬‬

‫‪1 0 0‬‬

‫‪١٢٦‬‬

‫‪In d e x‬‬

‫)‪z(t‬‬
‫‪56.3‬‬
‫‪56.9‬‬
‫‪56.0‬‬
‫‪59.4‬‬
‫‪68.2‬‬
‫‪69.8‬‬
‫‪67.9‬‬
‫‪58.2‬‬
‫‪52.7‬‬
‫‪53.3‬‬
‫‪53.2‬‬
‫‪51.5‬‬
‫‪52.3‬‬
‫‪53.9‬‬
‫‪57.9‬‬
‫‪66.4‬‬
‫‪74.8‬‬

‫ ﻧﺠﺪ‬wt = (1 − B ) zt ‫ﺑﺄﺧﺬ اﻟﻔﺮق اﻷول ﻹﺳﺘﻘﺮار اﻟﻤﺘﻮﺳﻂ‬

y(t)

1 0

0

-1 0

5 0

In d e x

1 0 0

1 5 0

‫وﻟﻬﺎ دوال ﺗﺮاﺑﻂ ذاﺗﻲ وﺗﺮاﺑﻂ ذاﺗﻲ ﺟﺰﺋﻲ‬
Autocorrelation

Autocorrelation Function for y(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

Lag

Corr

T

LBQ

Lag

Corr

T

LBQ

Lag

Corr

T

LBQ

Lag

Corr

T

LBQ

1
2
3
4
5
6
7
8
9
10
11
12

0.47
-0.02
-0.37
-0.28
-0.21
-0.19
-0.21
-0.27
-0.32
-0.00
0.46
0.86

6.32
-0.22
-4.10
-2.82
-2.00
-1.77
-1.92
-2.46
-2.85
-0.02
3.91
6.71

40.60
40.67
65.83
80.10
87.92
94.41
102.29
115.89
135.47
135.48
176.37
318.37

13
14
15
16
17
18
19
20
21
22
23
24

0.43
-0.01
-0.33
-0.25
-0.19
-0.17
-0.19
-0.25
-0.31
-0.01
0.42
0.79

2.76
-0.03
-2.00
-1.49
-1.12
-0.97
-1.09
-1.42
-1.76
-0.06
2.35
4.28

354.69
354.69
375.77
388.04
395.15
400.68
407.71
419.89
439.21
439.24
475.10
602.75

25
26
27
28
29
30
31
32
33
34
35
36

0.41
-0.00
-0.29
-0.23
-0.17
-0.16
-0.17
-0.23
-0.28
-0.01
0.37
0.72

2.06
-0.01
-1.43
-1.08
-0.82
-0.78
-0.83
-1.06
-1.31
-0.06
1.70
3.27

638.60
638.60
656.89
667.77
674.14
679.93
686.56
697.63
714.89
714.93
745.11
860.42

37
38
39
40
41
42
43
44

0.40
0.02
-0.27
-0.21
-0.17
-0.15
-0.16
-0.20

1.72
0.07
-1.14
-0.89
-0.71
-0.61
-0.65
-0.83

896.48
896.54
913.29
923.75
930.47
935.55
941.37
950.82

2

12

22

32

42

Partial Autocorrelation

Partial Autocorrelation Function for y(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

2

12

Lag PAC

T

0.47
-0.32
-0.30
0.09
-0.23
-0.25
-0.16
-0.43
-0.64
-0.31
-0.16
0.41

6.32
-4.21
-4.00
1.22
-3.05
-3.33
-2.09
-5.67
-8.56
-4.11
-2.18
5.42

1
2
3
4
5
6
7
8
9
10
11
12

Lag PAC
13
14
15
16
17
18
19
20
21
22
23
24

-0.27
-0.01
0.01
-0.07
-0.01
0.02
0.03
0.07
-0.12
-0.02
-0.06
0.13

22
T
-3.62
-0.17
0.13
-0.95
-0.16
0.24
0.44
0.89
-1.62
-0.26
-0.86
1.74

32

Lag PAC
25
26
27
28
29
30
31
32
33
34
35
36

-0.02
-0.06
0.07
-0.00
-0.02
-0.04
0.01
0.03
-0.02
0.00
-0.09
0.05

T
-0.33
-0.74
0.99
-0.05
-0.26
-0.60
0.20
0.37
-0.25
0.06
-1.17
0.61

42

Lag PAC
37
38
39
40
41
42
43
44

0.03
-0.03
-0.01
0.02
-0.04
0.03
-0.00
0.03

T
0.38
-0.35
-0.17
0.23
-0.56
0.35
-0.00
0.43

‫ وﻳﻨﺘﺞ اﻟﻤﺘﺴﻠﺴﻠﺔ‬wt = (1 − B12 ) (1 − B ) zt ‫ أي‬١٢ ‫ﻧﺮى اﻧﻬﺎ ﺗﺤﺘﺎج إﻟﻰ ﺗﻔﺮﻳﻖ ﻣﻮﺳﻤﻲ ﻣﻦ اﻟﺮﺗﺒﺔ‬
‫اﻟﺘﺎﻟﻴﺔ‬

١٢٧

w(t)

5

0

-5
In d e x

50

100

150

‫وﻟﻬﺎ دوال ﺗﺮاﺑﻂ ذاﺗﻲ وﺗﺮاﺑﻂ ذاﺗﻲ ﺟﺰﺋﻲ‬
Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r w ( t )
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
- 0 .2
- 0 .4
- 0 .6
- 0 .8
- 1 .0

10

20

30

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

1
2
3
4
5
6
7
8
9
10
11
12

-0 .0 1
-0 .1 9
-0 .2 4
-0 .0 0
0 .0 0
-0 .0 4
-0 .0 4
-0 .0 3
0 .2 2
0 .1 0
0 .0 8
-0 .4 2

- 0 .1 9
- 2 .4 8
- 2 .9 7
- 0 .0 3
0 .0 3
- 0 .5 0
- 0 .4 9
- 0 .3 1
2 .5 6
1 .1 6
0 .9 3
- 4 .6 3

0 .0 4
6 .3 1
1 6 .1 0
1 6 .1 0
1 6 .1 0
1 6 .4 1
1 6 .7 1
1 6 .8 3
2 5 .2 6
2 7 .1 3
2 8 .3 7
5 9 .4 0

13
14
15
16
17
18
19
20
21
22
23
24

-0 .0 8
0 .1 0
0 .1 2
-0 .0 2
-0 .0 3
0 .1 0
0 .0 7
-0 .0 1
-0 .2 0
-0 .0 2
0 .1 0
0 .1 4

-0 .8 4
0 .9 8
1 .1 5
-0 .1 7
-0 .2 8
1 .0 1
0 .7 0
-0 .0 5
-1 .9 4
-0 .2 3
0 .9 5
1 .3 0

6 0 .6 9
6 2 .5 0
6 5 .0 1
6 5 .0 7
6 5 .2 2
6 7 .2 4
6 8 .2 3
6 8 .2 4
7 5 .9 4
7 6 .0 5
7 8 .0 2
8 1 .7 8

25
26
27
28
29
30
31
32
33
34
35
36

- 0 .0 8
- 0 .1 3
- 0 .0 1
0 .0 6
0 .1 2
- 0 .1 5
- 0 .0 8
- 0 .0 3
0 .1 9
0 .1 7
- 0 .1 0
- 0 .2 2

-0 .7 6
-1 .1 8
-0 .1 3
0 .5 6
1 .0 9
-1 .3 6
-0 .7 2
-0 .2 8
1 .6 6
1 .4 9
-0 .8 6
-1 .9 2

8 3 .0 9
8 6 .2 8
8 6 .3 3
8 7 .0 7
8 9 .9 4
9 4 .4 7
9 5 .8 0
9 6 .0 1
1 0 3 .2 0
1 0 9 .1 9
1 1 1 .2 4
1 2 1 .7 7

40

L ag

T

LBQ

3 7 0 .0 5 0 .4 6
3 8 0 .1 0 0 .8 4
3 9 0 .0 2 0 .1 5
4 0 -0 .0 2 - 0 .1 6
4 1 -0 .1 2 - 1 .0 1

C o rr

1 2 2 .4 0
1 2 4 .5 2
1 2 4 .5 9
1 2 4 .6 7
1 2 7 .8 4

Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r w ( t )
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

10

20

30

40

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

1
2
3
4
5
6
7
8
9
10
11
12

- 0 .0 1
- 0 .1 9
- 0 .2 6
- 0 .0 7
- 0 .1 1
- 0 .1 4
- 0 .1 1
- 0 .1 3
0 .1 4
0 .0 7
0 .1 7
- 0 .3 2

- 0 .1 9
- 2 .4 8
- 3 .2 8
- 0 .8 5
- 1 .4 3
- 1 .8 4
- 1 .4 6
- 1 .6 6
1 .7 6
0 .9 5
2 .1 9
- 4 .0 8

13
14
15
16
17
18
19
20
21
22
23
24

-0 .0 5
0 .0 1
-0 .0 4
-0 .0 3
-0 .0 2
0 .0 6
0 .0 5
-0 .0 3
-0 .0 5
0 .0 5
0 .1 9
0 .0 1

- 0 .6 9
0 .1 5
- 0 .5 0
- 0 .3 3
- 0 .2 3
0 .7 8
0 .6 2
- 0 .4 2
- 0 .6 8
0 .6 3
2 .5 0
0 .0 9

25
26
27
28
29
30
31
32
33
34
35
36

- 0 .1 2
- 0 .0 8
- 0 .0 3
- 0 .0 3
0 .0 7
- 0 .1 1
- 0 .0 2
- 0 .1 0
- 0 .0 1
0 .1 8
0 .0 7
- 0 .0 9

-1 .4 9
-0 .9 7
-0 .3 2
-0 .3 6
0 .8 7
-1 .3 9
-0 .2 9
-1 .2 9
-0 .0 9
2 .3 1
0 .9 5
-1 .1 8

37
38
39
40
41

0 .0 2
- 0 .0 8
0 .0 3
0 .0 5
- 0 .0 3

0 .2 4
- 1 .0 6
0 .4 4
0 .6 3
- 0 .3 6

‫ﻣﻦ اﻷﻧﻤﺎط اﻟﻤﺸﺎهﺪة ﻟﺪوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻗﺪ ﻳﻜﻮن اﻟﻨﻤﻮذج اﻟﻤﻨﺎﺳﺐ هﻮ‬
‫ أي‬SARIMA(١،١،١)(٠،١،١)١٢

١٢٨

(1 − φ B ) (1 − B12 ) (1 − B ) zt = (1 − θ B ) (1 − ΘB12 ) at
:‫ﻧﻄﺒﻖ هﺬا اﻟﻨﻤﻮذج ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ آﺎﻟﺘﺎﻟﻲ‬
MTB > ARIMA 1 1 1 0 1 1 12 'z(t)' 'RESI2';
SUBC>
NoConstant;
SUBC> GACF;
SUBC> GPACF;
SUBC> GHistogram;
SUBC> GNormalplot.

ARIMA Model
ARIMA model for z(t)
Estimates at each iteration
Iteration
SSE
Parameters
0
307.653
0.100
0.100
1
281.217
0.100
0.100
2
262.275
0.226
0.231
3
262.027
0.376
0.381
4
261.770
0.526
0.531
5
261.426
0.675
0.681
6
260.905
0.824
0.831
7
260.036
0.970
0.981
8
227.926
0.835
0.980
9
221.838
0.748
0.980
10
221.665
0.738
0.980
11
221.637
0.738
0.980
12
221.610
0.737
0.980
13
221.585
0.737
0.980
Relative change in each estimate less than
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.7374
0.0620
MA
1
0.9796
0.0017
SMA 12
0.5898
0.0736

0.100
0.250
0.400
0.401
0.401
0.402
0.403
0.405
0.536
0.576
0.586
0.589
0.589
0.590
0.0010

T
11.89
582.86
8.01

Differencing: 1 regular, 1 seasonal of order 12
Number of observations: Original series 178, after differencing 165
Residuals:
SS = 214.393 (backforecasts excluded)
MS =
1.323 DF = 162
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
15.7(DF= 9)
30.9(DF=21)
61.6(DF=33)

48
67.1(DF=45)

‫أي أن اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻟﻬﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ هﻮ‬

(1 − 0.74 B ) (1 − B ) (1 − B ) zt = (1 − 0.98B ) (1 − 0.59 B12 ) at ,
12

at

N ( 0,1.323)

‫ﻻﺣﻆ ان‬

φ = 0.74, s.e. (φ ) = 0.062, t = 11.89
θ = 0.96, s.e. (θ ) = 0.0017, t = 582.86
Θ = 0.59, s.e. ( Θ ) = 0.074, t = 8.01

.‫أي ان اﻟﻤﻌﺎﻟﻢ ﻋﺎﻟﻴﺔ اﻟﻤﻌﻨﻮﻳﺔ‬
:‫ﻓﺤﺺ اﻟﺒﻮاﻗﻲ‬
‫إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ‬
MTB > ZTest 0.0 1.15 'RESI1';
SUBC>
Alternative 0.

Z-Test

١٢٩

Test of mu = 0.0000 vs mu not = 0.0000
The assumed sigma = 1.15
Variable
RESI1

N
165

Mean
-0.0144

StDev
1.1433

SE Mean
0.0895

Z
-0.16

P
0.87

‫ اي ﻻﻧﺮﻓﺾ أن ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ ﺻﻔﺮا‬٠٫٠٥ ‫ وهﻲ اآﺒﺮ ﻣﻦ‬P-value=٠٫٨٧ ‫ﻻﺣﻆ ان اﻟـ‬
‫إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬
MTB > Runs 0 'RESI1'.

Runs Test
RESI1
K =

0.0000

The observed number of runs = 67
The expected number of runs = 82.4061
73 Observations above K
92 below
The test is significant at 0.0149

‫ اي اﻧﻨﺎ ﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ وهﺬا ﻳﺤﺘﺎج إﻟﻰ إﺟﺮاء‬٠٫٠٥ ‫اﻹﺧﺘﺒﺎر ﻏﻴﺮﻣﻌﻨﻮي ﻋﻨﺪ‬
:‫ ﻋﻠﻰ اﻟﻮﺳﻴﻂ اﻟﺘﺎﻟﻲ‬Sign Test ‫إﺧﺘﺒﺎر ﺁﺧﺮ أآﺜﺮ ﻗﻮة ﻣﻦ إﺧﺘﺒﺎر اﻟﺠﺮي ﻣﺜﻞ إﺧﺘﺒﺎر اﻹﺷﺎرة‬
MTB > STest 0.0 'RESI1';
SUBC>
Alternative 0.

Sign Test for Median
Sign test of median = 0.00000 versus
RESI1

N
165

N*
13

Below
92

Equal
0

not =
Above
73

0.00000
P
0.1611

Median
-0.08139

‫ وﻟﻠﺘﺄآﺪ ﻧﺠﺮي إﺧﺘﺒﺎر وﻟﻜﻮآﺴﻮن ﻹﺷﺎرات اﻟﺮﺗﺐ ﻋﻠﻰ اﻟﻮﺳﻴﻂ‬٠٫١٦١١ ‫واﻹﺧﺘﺒﺎر ﻣﻌﻨﻮي ﻋﻨﺪ‬
‫اﻟﺘﺎﻟﻲ‬
MTB > WTest 0.0 'RESI1';
SUBC>
Alternative 0.

Wilcoxon Signed Rank Test
Test of median = 0.000000 versus median not = 0.000000

RESI1

N
165

Number
Missing
13

N for
Test
165

Wilcoxon
Statistic
6321.0

P
0.392

Estimated
Median
-0.05940

٠٫٣٩٢ ‫واﻹﺧﺘﺒﺎر اﻳﻀﺎ ﻣﻌﻨﻮي ﻋﻨﺪ‬
:‫إﺧﺘﺒﺎر إﺳﺘﻘﻼل اﻟﺒﻮاﻗﻲ‬
‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬

١٣٠

A C F o f R e s id u a ls f o r z ( t)
( w it h 9 5 % c o n f id e n c e l im it s f o r t h e a u t o c o r r e l a t io n s )
1 .0
0 .8

Autocorrelation

0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

3

6

9

12

15

18

21

24

27

30

33

36

39

36

39

Lag

P A C F o f R e s id u a ls f o r z ( t)
( w it h 9 5 % c o n f id e n c e l im it s f o r t h e p a r t ia l a u t o c o r r e l a t io n s )
1 .0

Partial Autocorrelation

0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

3

6

9

12

15

18

21

24

27

30

33

Lag

.‫ﻧﻼﺣﻆ اﻧﻬﺎ ﺗﻌﻄﻲ اﻧﻤﺎط اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء أي اﻧﻬﺎ ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ وإذا آﺎﻧﺖ ﻃﺒﻴﻌﻴﺔ ﻓﻬﻲ ﻣﺴﺘﻘﻠﺔ‬
:‫إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬
Histogram of the Residuals
(response is z(t))
50

Frequency

40

30

20

10

0
-6

-5

-4

-3

-2

-1

0

1

2

Residual

١٣١

3

4

Normal Probability Plot of the Residuals
(response is z(t))
4
3
2

Residual

1
0
-1
-2
-3
-4
-5
-3

-2

-1

0

1

2

3

Normal Score

K-S Test for Residuals

.999

Probability

.99
.95
.80
.50
.20
.05
.01
.001
-5

-4

-3

-2

-1

0

1

2

3

RESI1
Average: -0.0144171
StDev: 1.14327
N: 165

Kolmogorov-Smirnov Normality Test
D+: 0.117 D-: 0.140 D : 0.140
Approximate P-Value < 0.01

‫ اي‬α = 0.05 ‫ اذا اﻹﺧﺘﺒﺎر ﻣﻌﻨﻮي ﻋﻨﺪ‬٠٫٠١ ‫ أﻗﻞ ﻣﻦ‬K-S ‫ ﻹﺧﺘﺒﺎر‬P-value ‫ﻻﺣﻆ ان اﻟـ‬
.‫ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬
:‫اﻟﺘﻨﺒﺆ ﺑﺈﺳﺘﺨﺪام اﻟﻨﻤﻮذج‬
‫ ﻓﺘﺮات ﺗﻨﺒﺆ‬٩٥% ‫ ﻗﻴﻤﺔ ﻣﺴﺘﻘﺒﻠﻴﺔ ﻣﻊ‬٣٦ ‫ﺳﻨﻘﻮم ﺑﺎﻟﺘﻨﺒﺆ ﻋﻦ‬
Forecasts from period 178
Period
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193

Forecast
57.7885
55.8516
54.7429
54.1820
54.6298
54.9152
55.6640
59.3778
68.6924
74.0698
73.9650
63.5866
58.9095
56.7768
55.5237

95 Percent Limits
Lower
Upper
55.5332
60.0437
53.0220
58.6812
51.6264
57.8594
50.9063
57.4578
51.2601
57.9994
51.4875
58.3430
52.1986
59.1294
55.8869
62.8688
65.1833
72.2015
70.5472
77.5925
70.4317
77.4983
60.0447
67.1286
55.1843
62.6347
52.9407
60.6128
51.6169
59.4305

١٣٢

Actual

‫‪50.9022‬‬
‫‪51.2380‬‬
‫‪51.4409‬‬
‫‪52.1278‬‬
‫‪55.7947‬‬
‫‪65.0729‬‬
‫‪70.4217‬‬
‫‪70.2940‬‬
‫‪59.8968‬‬
‫‪55.0418‬‬
‫‪52.7962‬‬
‫‪51.4676‬‬
‫‪50.7472‬‬
‫‪51.0774‬‬
‫‪51.2749‬‬
‫‪51.9568‬‬
‫‪55.6189‬‬
‫‪64.8927‬‬
‫‪70.2374‬‬
‫‪70.1057‬‬
‫‪59.7046‬‬

‫‪58.8105‬‬
‫‪59.2131‬‬
‫‪59.4653‬‬
‫‪60.1905‬‬
‫‪63.8885‬‬
‫‪73.1929‬‬
‫‪78.5647‬‬
‫‪78.4575‬‬
‫‪68.0793‬‬
‫‪63.5664‬‬
‫‪61.5364‬‬
‫‪60.3514‬‬
‫‪59.7315‬‬
‫‪60.1357‬‬
‫‪60.3903‬‬
‫‪61.1184‬‬
‫‪64.8194‬‬
‫‪74.1270‬‬
‫‪79.5021‬‬
‫‪79.3983‬‬
‫‪69.0235‬‬

‫‪194‬‬
‫‪195‬‬
‫‪196‬‬
‫‪197‬‬
‫‪198‬‬
‫‪199‬‬
‫‪200‬‬
‫‪201‬‬
‫‪202‬‬
‫‪203‬‬
‫‪204‬‬
‫‪205‬‬
‫‪206‬‬
‫‪207‬‬
‫‪208‬‬
‫‪209‬‬
‫‪210‬‬
‫‪211‬‬
‫‪212‬‬
‫‪213‬‬
‫‪214‬‬

‫‪54.8564‬‬
‫‪55.2256‬‬
‫‪55.4531‬‬
‫‪56.1592‬‬
‫‪59.8416‬‬
‫‪69.1329‬‬
‫‪74.4932‬‬
‫‪74.3758‬‬
‫‪63.9881‬‬
‫‪59.3041‬‬
‫‪57.1663‬‬
‫‪55.9095‬‬
‫‪55.2394‬‬
‫‪55.6066‬‬
‫‪55.8326‬‬
‫‪56.5376‬‬
‫‪60.2191‬‬
‫‪69.5099‬‬
‫‪74.8697‬‬
‫‪74.7520‬‬
‫‪64.3640‬‬

‫وﻧﺮﺳﻤﻬﺎ ﺑﺎﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪8 0‬‬

‫‪Forecast‬‬

‫‪7 0‬‬

‫‪6 0‬‬

‫‪5 0‬‬

‫‪3 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪0‬‬

‫‪1 0‬‬

‫‪T im e‬‬

‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﺑﻜﺎﻣﻠﻬﺎ ﻣﻊ اﻟﺘﻨﺒﺆات وﻓﺘﺮات اﻟﺘﻨﺒﺆ‬

‫‪80‬‬

‫‪60‬‬

‫‪50‬‬

‫‪200‬‬

‫‪100‬‬

‫‪T im e‬‬

‫‪١٣٣‬‬

‫‪0‬‬

‫‪Forecast‬‬

‫‪70‬‬

١٣٤

‫اﻟﻔﺼﻞ اﻟﺴﺎﺑﻊ‬
‫ورﻗﺔ ﺗﺪرﻳﺐ ﻋﻤﻠﻲ ﻋﻠﻲ اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ ﻧﻤﺎذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪-‬اﻹﻧﺤﺪار‬
‫اﻟﺬاﺗﻲ‬
‫‪Forecasting By ARMA Models‬‬
‫اﻟﻤﺸﺎهﺪات اﻟﺘﺎﻟﻴﺔ ﻟﻈﺎهﺮة ﻋﺸﻮاﺋﻴﺔ ﻣﺴﺠﻠﺔ ﻋﻠﻲ ﺷﻜﻞ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ‬
‫‪28.0‬‬
‫‪28.0‬‬
‫‪29.5‬‬

‫‪23.5‬‬
‫‪30.5‬‬
‫‪32.3‬‬

‫‪21.3‬‬
‫‪29.6‬‬
‫‪30.6‬‬

‫‪24.5‬‬
‫‪36.5‬‬
‫‪26.5‬‬

‫‪23.5‬‬
‫‪36.5‬‬
‫‪20.7‬‬
‫‪16.4‬‬

‫‪15.3‬‬
‫‪25.0‬‬
‫‪16.0‬‬
‫‪23.4‬‬

‫‪15.5‬‬
‫‪25.3‬‬
‫‪19.0‬‬
‫‪26.4‬‬

‫‪21.0‬‬
‫‪17.3‬‬
‫‪19.7‬‬
‫‪32.2‬‬

‫‪12.0‬‬
‫‪24.0‬‬
‫‪26.0‬‬
‫‪28.3‬‬

‫‪20.5‬‬
‫‪15.5‬‬
‫‪21.5‬‬
‫‪31.3‬‬

‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬
‫;‪MTB > TSPlot C1‬‬
‫>‪SUBC‬‬
‫;‪Index‬‬
‫>‪SUBC‬‬
‫;‪TDisplay 11‬‬
‫>‪SUBC‬‬
‫;‪Symbol‬‬
‫>‪SUBC‬‬
‫‪Connect.‬‬

‫‪30‬‬

‫‪C‬‬
‫‪1‬‬
‫‪20‬‬

‫‪10‬‬
‫‪30‬‬

‫‪10‬‬

‫‪20‬‬

‫اوﻻ ﻧﻮﺟﺪ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬

‫‪١٣٥‬‬

‫‪Index‬‬

MTB > %ACF C1.

Autocorrelation

Autocorrelation Function for C1
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

1

2

3

4

Lag

Corr

T

LBQ

1
2
3
4
5
6
7

0.63
0.30
0.14
-0.05
-0.24
-0.30
-0.24

3.79
1.35
0.59
-0.20
-1.01
-1.22
-0.94

15.62
19.27
20.07
20.17
22.71
26.71
29.40

5

Lag

6

Corr

7

T

LBQ

8 -0.20 -0.78
9 -0.12 -0.44

31.42
32.10

8

9

MTB > %PACF C1.
Executing from file: E:\MTBWIN\MACROS\PACF.MAC

Partial Autocorrelation

Partial Autocorrelation Function for C1
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

1

2

3

4

Lag PAC
1
2
3
4
5
6
7

0.63
-0.16
0.04
-0.20
-0.18
-0.04
0.02

5

6

T

Lag PAC

T

3.79
-0.98
0.22
-1.20
-1.09
-0.23
0.12

8 -0.07
9 0.05

-0.39
0.29

7

8

9

ARMA (1,1) ‫ﻧﻼﺣﻆ ﻣﻦ اﻧﻤﺎط اﻟﺪاﻟﺘﻴﻦ ان اﻟﻤﺸﺎهﺪات ﻗﺪ ﺗﻜﻮن ﻣﻦ ﻧﻤﻮذج‬
‫ﻧﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح‬
MTB > Name c17 = 'RESI1'
MTB > ARIMA 1 0 1 C1 'RESI1';
SUBC>
Constant;
SUBC>
Forecast 5 c14 c15 c16;
SUBC> GACF;
SUBC> GPACF;
SUBC> GNormalplot.

١٣٦

ARIMA Model
ARIMA model for C1
Estimates at each iteration
Iteration
SSE
Parameters
0
1337.71
0.100
0.100
21.918
1
936.95
0.250
-0.049
18.193
2
849.78
0.211
-0.199
19.106
3
751.53
0.215
-0.349
18.941
4
658.66
0.266
-0.499
17.594
5
592.30
0.372
-0.649
14.890
6
580.80
0.433
-0.699
13.314
7
579.30
0.455
-0.714
12.698
8
579.11
0.464
-0.719
12.470
9
579.08
0.467
-0.721
12.386
10
579.08
0.468
-0.722
12.356
11
579.08
0.468
-0.722
12.345
Relative change in each estimate less than 0.0010
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.4684
0.1755
MA
1
-0.7221
0.1380
Constant
12.345
1.154
Mean
23.221
2.170

T
2.67
-5.23
10.70

Number of observations: 36
Residuals:
SS = 523.365 (backforecasts excluded)
MS = 15.860 DF = 33
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
48
Chi-Square 7.2(DF=10)
15.9(DF=22) * (DF= *)
* (DF= *)
Forecasts from period 36
Period
37
38
39
40
41

Forecast
14.7649
19.2606
21.3663
22.3524
22.8143

95 Percent Limits
Lower
Upper
6.9578
22.5720
7.1228
31.3985
8.4715
34.2610
9.2975
35.4074
9.7245
35.9041

zt = 12.345 + 0.4684 zt −1 + at − 0.7221at −1 , at
‫ﺣﻴﺚ‬
φˆ1 = 0.4684 se φˆ1 = 0.1755 t = 2.67

Actual

:‫اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬
WN ( 0,15.86 ) ∀t

( )
θˆ = −0.7221 se (θˆ ) = 0.1380 t = −5.23
δˆ = 12.345 se (δˆ ) = 1.154 t = 10.70
1

1

σˆ 2 = 15.86 df = 33
‫ ﻓﻤﺜﻼ اﻟﻔﺮﺿﻴﺔ‬α = 0.05 ‫وﻧﻼﺣﻆ ان ﺟﻤﻴﻊ اﻟﻤﻘﺪرات ﻣﻌﻨﻮﻳﺔ ﻋﻨﺪ‬

١٣٧

H 0 : φ1 = 0
H1 : φ1 ≠ 0
‫ أي‬α = 0.05 ‫ وهﻲ ﻣﻌﻨﻮﻳﺔ ﻋﻨﺪ‬t =

0.4684
φˆ1
=
= 2.6689 ‫ﻧﺨﺘﺒﺮهﺎ ﺑﺎﻹﺣﺼﺎﺋﺔ‬
0.1755
se φˆ1

( )

‫ وﺑﺎﻟﻤﺜﻞ ﻟﺠﻤﻴﻊ اﻟﻘﺪرات اﻻﺧﺮى‬φ1 = 0 ‫اﻧﻨﺎ ﻧﺮﻓﺾ ان‬

‫ﺛﺎﻧﻴﺎ ﻓﺤﺺ اﻟﺒﻮاﻗﻲ‬
:‫إﺧﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬
‫اﻹﺧﺘﺒﺎر هﻮ‬

H 0 : µa = 0, H1 : µa ≠ 0

MTB > TTest 0.0 'RESI1';
SUBC>
Alternative 0.

T-Test of the Mean
Test of mu = 0.000 vs mu not = 0.000
Variable
RESI1

N
36

Mean
0.344

StDev
3.851

SE Mean
0.642

T
0.54

P
0.60

‫ أي ان اﻹﺧﺘﺒﺎر‬α = 0.05 ‫ وهﻲ اآﺒﺮ ﻣﻦ‬0.6 ‫ ﻟﻬﺎ هﻲ‬P-Value ‫ واﻟـ‬t = 0.54 ‫ﻻﺣﻆ ان‬
‫ﻏﻴﺮ ﻣﻌﻨﻮي أي ﻳﻤﻜﻦ إﻋﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ ﻳﺴﺎوي اﻟﺼﻔﺮ‬
:‫إﺧﺘﺒﺎرﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬
‫وﻧﺴﺘﺨﺪم ﻟﺬﻟﻚ إﺧﺘﺒﺎر اﻟﺠﺮي‬Runs Test
MTB > Runs 'RESI1'.

Runs Test
RESI1
K =

0.3443

The observed number of runs = 21
The expected number of runs = 19.0000
18 Observations above K
18 below
The test is significant at 0.4989
Cannot reject at alpha = 0.05

α = 0.05 ‫ﻻﻳﻤﻜﻨﻨﺎ رﻓﺾ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ ﻋﻨﺪ‬
:‫إﺧﺘﺒﺎر ﺗﺮاﺑﻂ اﻟﺒﻮاﻗﻲ‬
‫وﻧﺴﺘﺨﺪم ﻟﺬﻟﻚ إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬

ACF of Residuals for C1
(with 95% confidence limits for the autocorrelations)
1.0
0.8

orrelation

0.6
0.4
0.2
0.0

١٣٨

PACF of Residuals for C1
(with 95% confidence limits for the partial autocorrelations)
1.0

Partial Autocorrelation

0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
1

2

3

4

5

6

7

8

9

Lag

‫ﻧﻼﺣﻆ اﻧﻪ ﻻﻳﻮﺟﺪ أي ﺗﺮاﺑﻂ ﻣﻦ أي درﺟﺔ ﺑﻴﻦ اﻟﻘﻴﻢ اﻟﻤﺨﺘﻠﻔﺔ ﻟﻠﺒﻮاﻗﻲ أي اﻧﻬﺎ ﺗﻈﻬﺮ اﻧﻤﺎط ﺗﺘﻤﺸﻲ‬
‫ﻣﻊ آﻮﻧﻬﺎ ﻣﺘﺴﻠﺴﺔ ﺿﺠﺔ ﺑﻴﻀﺎء‬
Normal Probability Plot ‫واﺧﻴﺮا ﻧﺨﺘﺒﺮ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ ﺑﺎﻟـ‬

Normal Probability Plot of the Residuals
(response is C1)

Residual

10

0

-10

-2

-1

0

Normal Score

١٣٩

1

2

( ‫وهﻮ ﻣﻘﺒﻮل ) ﻧﻮﻋﺎ‬
‫إذا ﻳﻤﻜﻦ إﻋﺘﺒﺎر اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻣﻨﺎﺳﺒﺎ‬
‫ ﺗﻨﺒﺆ ﻟﻬﺎ‬95% ‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﺘﻨﺒﺆات ﻟﺨﻤﺴﺔ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ وﻓﺘﺮات‬

MTB > TSPlot C14 C15 C16;
SUBC>
Index;
SUBC>
TDisplay 11;
SUBC>
Symbol;
SUBC>
Connect;
SUBC> overlay.

Time Series Plot for C1
(with forecasts and their 95% confidence limits)

36

C1

26

16

6

5

10

15

20

25

30

35

Time

:‫ﻣﺜﺎل ﺁﺧﺮ‬
‫اﻟﻤﺸﺎهﺪات اﻟﺘﺎﻟﻴﺔ ﻟﻈﺎهﺮة ﻋﺸﻮاﺋﻴﺔ ﻣﺴﺠﻠﺔ ﻋﻠﻲ ﺷﻜﻞ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ‬
10.38
11.68
9.10
10.01
9.51
9.48
7.23
10.85
6.89

11.86
11.17
9.09
9.37
9.24
7.38
8.42
10.41
5.96

10.97
10.53
9.35
8.69
8.66
6.90
9.61
9.96
6.80

10.80
10.01
8.82
8.19
8.86
6.94
9.05
9.61
7.68

9.79
9.91
9.32
8.67
8.05
6.24
9.26
8.76
8.38

10.39
9.14
9.01
9.55
7.79
6.84
9.22
8.18
8.52

10.42
9.16
9.00
8.92
6.75
6.85
9.38
7.21
9.74

10.82
9.55
9.80
8.09
6.75
6.90
9.10
7.13
9.31

11.40
9.67
9.83
9.37
7.82
7.79
7.95
9.10
9.89

11.32 11.44
8.44
8.24
9.72
9.89
10.13 10.14
8.64 10.58
8.18
7.51
8.12
9.75
8.25
7.91
9.96

MTB > TSPlot C10;
SUBC>
Index;
SUBC>
TDisplay 11;
SUBC>
Symbol;
SUBC>
Connect.

12
11

C
1
0

10
9
8

١٤٠

7
6
Index

10

20

30

40

50

60

70

80

90

‫ﻧﻔﺤﺺ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬

Autocorrelation

Autocorrelation Function for C10
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

2

Lag Corr
1
2
3
4
5
6
7

0.83
0.61
0.46
0.37
0.33
0.28
0.26

T

12

LBQ

8.24 69.92
3.91 107.90
2.56 129.56
1.95 143.87
1.65 155.04
1.40 163.68
1.28 171.23

Lag Corr
8
9
10
11
12
13
14

0.26
0.26
0.18
0.09
0.04
0.03
0.04

T

LBQ

22

Lag Corr

1.26 178.83
1.21 186.14
0.84 189.86
0.43 190.87
0.20 191.09
0.13 191.19
0.19 191.39

T

LBQ

15 0.05 0.21 191.63
16 0.04 0.16 191.78
17 0.00 0.02 191.78
18 -0.03 -0.15 191.91
19 -0.05 -0.24 192.26
20 -0.05 -0.24 192.60
21 0.01 0.07 192.63

Lag Corr

T

LBQ

22 0.10 0.47 194.02
23 0.18 0.81 198.12
24 0.20 0.89 203.24

Partial Autocorrelation

Partial Autocorrelation Function for C10
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

2

12

Lag PAC

T

Lag PAC

T

1 0.83
2 -0.27
3 0.13
4 0.03
5 0.06
6 -0.02
7 0.09

8.24
-2.64
1.29
0.34
0.61
-0.21
0.91

8 0.05
9 0.00
10 -0.20
11 0.02
12 0.01
13 0.01
14 0.03

0.45
0.03
-1.98
0.19
0.09
0.12
0.34

22

Lag PAC
15
16
17
18
19
20
21

-0.01
-0.03
-0.07
-0.03
0.06
0.02
0.21

T

Lag PAC

T

-0.15
-0.25
-0.73
-0.26
0.60
0.20
2.03

22 0.05
23 0.06
24 -0.07

0.51
0.59
-0.65

AR ( 2 ) ‫ﻧﻼﺣﻆ ان اﻷﻧﻤﺎط ﺗﻘﺘﺮح ﻧﻤﻮذج إﻧﺤﺪار ذاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ‬
‫ﻧﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح‬
MTB > Name c17 = 'RESI1'
MTB > ARIMA 2 0 0 C10 'RESI1';
SUBC>
Constant;

١٤١

SUBC>
SUBC>
SUBC>
SUBC>
SUBC>

Forecast 5 c14 c15 c16;
GSeries;
GACF;
GPACF;
GNormalplot.

ARIMA Model
ARIMA model for C10
Estimates at each iteration
Iteration
SSE
Parameters
0
126.398
0.100
0.100
1
103.515
0.250
0.043
2
84.535
0.400
-0.014
3
69.407
0.550
-0.071
4
58.132
0.700
-0.128
5
50.724
0.850
-0.184
6
47.212
1.000
-0.239
7
46.918
1.053
-0.256
8
46.916
1.054
-0.255
9
46.916
1.054
-0.255
10
46.916
1.054
-0.255
Relative change in each estimate less than
Final Estimates of Parameters
Type
Coef
StDev
AR
1
1.0542
0.0992
AR
2
-0.2547
0.0993
Constant
1.81360
0.07092
Mean
9.0480
0.3538

7.283
6.434
5.586
4.738
3.887
3.030
2.163
1.838
1.816
1.814
1.814
0.0010

T
10.63
-2.56
25.57

Number of observations: 98
Residuals:
SS = 46.7518 (backforecasts excluded)
MS = 0.4921 DF = 95
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
48
Chi-Square 7.2(DF=10) 13.7(DF=22) 21.3(DF=34) 28.8(DF=46)
Forecasts from period 98
Period
99
100
101
102
103

Forecast
9.7950
9.6033
9.4432
9.3232
9.2375

95 Percent Limits
Lower
Upper
8.4198
11.1703
7.6050
11.6016
7.1234
11.7629
6.8446
11.8018
6.6825
11.7925

Actual

‫اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬
zt = 1.8136 + 1.0542 zt −1 − 0.2547 zt −2 + at , at WN ( 0,0.4921) ∀t
‫ ﻟﻬﺎ هﻲ‬t ‫ﻣﻘﺪرات اﻟﻤﻌﺎﻟﻢ وإﻧﺤﺮاﻓﺎﺗﻬﺎ اﻟﻤﻌﻴﺎرﻳﺔ وﻗﻴﻢ‬
φˆ1 = 1.0542 se φˆ1 = 0.0992 t = 10.63

( )
φˆ = −0.2547 se (φˆ ) = 0.0993 t = −2.56
δˆ = 1.8136 se (δˆ ) = 0.07092 t = 25.57
2

2

σˆ 2 = 0.4921 df = 95

α = 0.05 ‫ﻧﻼﺣﻆ ان ﺟﻤﻴﻊ اﻟﻤﻘﺪرات ﻣﻌﻨﻮﻳﺔ ﻋﻨﺪ‬
١٤٢

‫ﻟﻜﻲ ﻧﻘﺒﻞ ﺑﻬﺬا اﻟﻨﻤﻮذج ﻋﻠﻲ اﻧﻪ ﻣﻨﺎﺳﺐ ﻟﻠﺘﻨﺒﺆ ﻧﺠﺮي إﺧﺘﺒﺎرات ﻋﻠﻲ اﻟﺒﻮاﻗﻲ‬
MTB > TTest 0.0 'RESI1';
SUBC>
Alternative 0.

T-Test of the Mean
Test of mu = 0.0000 vs mu not = 0.0000
Variable
RESI1

N
98

Mean
StDev
-0.0082 0.6942

SE Mean
0.0701

T
-0.12

P
0.91

‫ هﻲ إﺣﺘﻤﺎل ان اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ‬P-Value ‫واﺿﺢ ﺟﺪا ان اﻷﺧﺘﺒﺎر ﻏﻴﺮ ﻣﻌﻨﻮي ) ﻣﻼﺣﻈﺔ اﻟـ‬
( ‫ﺻﺤﻴﺤﺔ‬
‫ﻧﺨﺘﺒﺮ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬
MTB > Runs 'RESI1'.

Runs Test
RESI1
K =

-0.0082

The observed number of runs = 47
The expected number of runs = 49.9184
47 Observations above K
51 below
The test is significant at 0.5529
Cannot reject at alpha = 0.05

‫أي ﻻﻳﻤﻜﻨﻨﺎ رﻓﺾ ﻓﺮﺿﻴﺔ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬
‫ﻧﻔﺤﺺ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬

ACF of Residuals for C10
(with 95% confidence limits for the autocorrelations)
1.0
0.8

Autocorrelation

0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

2

4

6

8

10

12

14

Lag

١٤٣

16

18

20

22

24

PACF of Residuals for C10
(with 95% confidence limits for the partial autocorrelations)
1.0

Partial Autocorrelation

0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

2

4

6

8

10

12

14

16

18

20

22

24

Lag

‫واﺿﺢ ﺟﺪا اﻧﻤﺎط اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬
‫ﻳﺒﻘﻲ ﻓﺤﺺ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬
Normal Probability Plot of the Residuals
(response is C10)
2

Residual

1

0

-1

-2
-3

( at

-2

-1

0

1

Normal Score

2

3

IIDN ( 0,0.4921) ‫وﻧﺴﺘﻄﻴﻊ ان ﻧﻘﻮل ان اﻟﺒﻮاﻗﻲ ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ) أي‬
‫ ﺗﻨﺒﺆ ﻟﻬﺎ‬95% ‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﺘﻨﺒﺆات ﻟﺨﻤﺴﺔ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ وﻓﺘﺮات‬

MTB > TSPlot C14 C15 C16;
SUBC>
Index;
SUBC>
TDisplay 11;
SUBC>
Symbol;
SUBC>
Connect;
SUBC> overlay.

Time Series Plot for C10
(with forecasts and their 95% confidence limits)
12
11

C10

10
9
8
7
6

١٤٤

‫ﺗﻤﺮﻳﻦ‪ :‬ﻃﺒﻖ ﻋﻠﻲ اﻟﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ ﻧﻤﻮذج )‪ AR (1‬وﻗﺎرن ﺑﻴﻦ اﻟﻨﺘﺎﺋﺞ‬

‫‪١٤٥‬‬

١٤٦

‫اﻟﻔﺼﻞ اﻟﺜﺎﻣﻦ‬
‫ﻣﺜﺎل ﻋﻠﻰ ﺗﺤﻠﻴﻞ اﻟﺒﻮاﻗﻲ وﻣﻌﺎﻳﻴﺮ إﺧﺘﻴﺎر اﻟﻨﻤﻮذج اﻟﻤﻨﺎﺳﺐ‬
‫‪Example on Residual Analysis and Model Selection‬‬
‫‪: Criteria‬‬
‫ﻟﻘﺪ ﻋﺮﻓﻨﺎ ﺳﺎﺑﻘﺎ اﻟﺒﻮاﻗﻲ ﻋﻠﻲ اﻧﻬﺎ اﻟﻘﻴﻢ اﻟﻤﺸﺎهﺪة ﻧﺎﻗﺺ اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﻓﻤﻦ ﻣﺸﺎهﺪات ﻣﻌﻄﺎة‬
‫‪ z1 , z2 ,..., zn‬وﻧﻤﻮذج ﻣﻄﺒﻖ ﻳﻨﺘﺞ ﻟﺪﻳﻨﺎ ﻗﻴﻢ ﻣﻄﺒﻘﺔ ‪ zˆ1 , zˆ2 ,..., zˆn‬وﺗﻜﺘﺐ اﻟﺒﻮاﻗﻲ‪:‬‬
‫‪ei = zi − zˆi , i = 1, 2,..., n‬‬
‫واﻟﺒﻮاﻗﻲ هﻲ ﻣﻘﺪرات اﻷﺧﻄﺎء ﻓﻲ اﻟﻨﻤﻮذج أي ‪ aˆi = ei , i = 1, 2,..., n‬وﻟﻬﺬا ﻳﺠﺐ ان ﺗﺤﻘﻖ‬
‫اﻟﺸﺮوط اﻟﻤﻔﺮوﺿﺔ ﻋﻠﻲ اﻷﺧﻄﺎء ﻓﻲ هﺬا اﻟﻨﻤﻮذج واﻟﺘﻲ ﻣﻨﻬﺎ‪:‬‬
‫‪ -١‬ﻣﺘﻮﺳﻂ اﻷﺧﻄﺎء ﻳﺴﺎوي اﻟﺼﻔﺮ‬
‫‪ -٢‬اﻷﺧﻄﺎء ﻋﺸﻮاﺋﻴﺔ و ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ أو ﻣﺴﺘﻘﻠﺔ ) وﻓﻲ آﺜﻴﺮ ﻣﻦ اﻟﻨﻤﺎذج ﻧﻔﺘﺮض ان اﻷﺧﻄﺎء‬
‫ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ﻣﺴﺘﻘﻞ وﻣﺘﻄﺎﺑﻖ ﺑﻤﺘﻮﺳﻂ ﺻﻔﺮي وﺗﺒﺎﻳﻦ ‪ σ 2‬أي ) ‪( at IIDN ( 0, σ 2‬‬
‫ﻟﻬﺬا ﻓﺈﻧﻨﺎ ﻧﺠﺮي ﺗﺤﻠﻴﻼ وهﻮ ﻣﺠﻤﻮﻋﺔ ﻣﻦ اﻹﺧﺘﺒﺎرات ﻋﻠﻲ اﻟﺒﻮاﻗﻲ ﻟﻨﺮي ﻓﻴﻤﺎ إذا آﺎﻧﺖ ﺗﺤﻘﻖ هﺬﻩ‬
‫اﻟﺸﺮوط وﻓﻲ هﺬﻩ اﻟﺤﺎﻟﺔ ﻧﻌﺘﺒﺮ اﻟﻨﻤﻮذج اﻟﻤﻄﺒﻖ ﻣﻘﺒﻮﻻ أﻣﺎ إذا ﻓﺸﻞ اﺣﺪ هﺬﻩ اﻹﺧﺘﺒﺎرات ﻓﻴﺠﺐ‬
‫ﻋﻠﻴﻨﺎ إﻋﺎدة اﻟﻨﻈﺮ وإﻗﺘﺮاح ﻧﻤﻮذج ﺁﺧﺮ‬
‫أوﻻ‪ :‬إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ‬
‫‪H 0 : E ( at ) = 0‬‬
‫‪H 1 : E ( at ) ≠ 0‬‬

‫وهﻮ إﺧﺘﺒﺎر ﺑﺬﻳﻠﻴﻦ وﻧﺴﺘﺨﺪم ﻓﻴﺔ اﻹﺣﺼﺎﺋﺔ‬

‫‪e‬‬

‫) ‪se ( e‬‬

‫= ‪ u‬واﻟﺘﻲ ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ﻗﻴﺎﺳﻲ ﻓﻌﻨﺪ‬

‫ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬ﻧﻌﺘﺒﺮ ان ‪ E ( at ) = 0‬إذا آﺎﻧﺖ ‪ ) u < 1.96‬هﺬا ﻋﻠﻲ إﻋﺘﺒﺎر ان‬
‫ﺣﺠﻢ اﻟﻌﻴﻨﺔ اآﺒﺮ ﻣﻦ ‪ ٣٠‬وﺣﺪة وهﺬا داﺋﻤﺎ ﻣﺘﺤﻘﻖ ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﺘﻲ ﻧﺪرﺳﻬﺎ (‬

‫ﻣﺜﺎل‪:‬‬
‫ﺳﻮف ﻧﻌﻮد اﻟﻲ ﻣﺜﺎل ﺗﻄﺒﻴﻖ ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻟﺜﺔ ﻋﻠﻲ اﻟﻤﺘﻐﻴﺮ ‪Metals‬‬
‫‪MTB > RETR 'E:\Mtbwin\DATA\EMPLOY.MTW'.‬‬
‫‪Retrieving worksheet from file: E:\Mtbwin\DATA\EMPLOY.MTW‬‬
‫‪Worksheet was saved on 6/ 5/1996‬‬
‫;'‪MTB > TSPlot 'Metals‬‬
‫>‪SUBC‬‬
‫;‪Index‬‬
‫>‪SUBC‬‬
‫;‪TDisplay 11‬‬
‫>‪SUBC‬‬
‫;‪Symbol‬‬
‫>‪SUBC‬‬
‫‪Connect.‬‬

‫‪50‬‬

‫‪M‬‬
‫‪e‬‬
‫‪ta‬‬
‫‪ls‬‬

‫‪45‬‬

‫‪40‬‬
‫‪60‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪Index‬‬

‫'‪MTB > Name c4 = 'AVER1' c5 = 'FITS1' c6 = 'RESI1‬‬

‫‪١٤٧‬‬

MTB > %MA 'Metals' 3;
SUBC>
Averages 'AVER1';
SUBC>
Fits 'FITS1';
SUBC>
Residuals 'RESI1'.
Executing from file: E:\MTBWIN\MACROS\MA.MAC

Moving average
Data
Length
NMissing

Metals
60.0000
0

Moving Average
Length: 3
Accuracy Measures
MAPE: 1.55036
MAD: 0.70292
MSD: 0.76433

Moving Average

Actual

Predicted

M
etals

50

Actual
Predicted

45
Moving Average
Length:

40
0

10

20

30

40

50

3

MAPE:

1.55036

MAD:

0.70292

MSD:

0.76433

60

Time

‫ﻻﺣﻆ اﻧﻨﺎ ﺧﺰﻧﺎ اﻟﺒﻮاﻗﻲ ﻓﻲ اﻟﻌﻤﻮد اﻟﺴﺎدس واﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﻓﻲ اﻟﻌﻤﻮد اﻟﺨﺎﻣﺲ‬
MTB > print c3 c6 c5

Data Display

Row

Metals

RESI1

FITS1

1
2
3
4
5
6
7
8
9

44.2
44.3
44.4
43.4
42.8
44.3
44.4
44.8
44.4

*
*
*
-0.90000
-1.23333
0.76667
0.90000
0.96667
-0.10000

*
*
*
44.3000
44.0333
43.5333
43.5000
43.8333
44.5000

١٤٨

10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

43.1
42.6
42.4
42.2
41.8
40.1
42.0
42.4
43.1
42.4
43.1
43.2
42.8
43.0
42.8
42.5
42.6
42.3
42.9
43.6
44.7
44.5
45.0
44.8
44.9
45.2
45.2
45.0
45.5
46.2
46.8
47.5
48.3
48.3
49.1
48.9
49.4
50.0
50.0
49.6
49.9
49.6
50.7
50.7
50.9
50.5
51.2
50.7
50.3
49.2
48.1

-1.43333
-1.50000
-0.96667
-0.50000
-0.60000
-2.03333
0.63333
1.10000
1.60000
-0.10000
0.46667
0.33333
-0.10000
-0.03333
-0.20000
-0.36667
-0.16667
-0.33333
0.43333
1.00000
1.76667
0.76667
0.73333
0.06667
0.13333
0.30000
0.23333
-0.10000
0.36667
0.96667
1.23333
1.33333
1.46667
0.76667
1.06667
0.33333
0.63333
0.86667
0.56667
-0.20000
0.03333
-0.23333
1.00000
0.63333
0.56667
-0.26667
0.50000
-0.16667
-0.50000
-1.53333
-1.96667

44.5333
44.1000
43.3667
42.7000
42.4000
42.1333
41.3667
41.3000
41.5000
42.5000
42.6333
42.8667
42.9000
43.0333
43.0000
42.8667
42.7667
42.6333
42.4667
42.6000
42.9333
43.7333
44.2667
44.7333
44.7667
44.9000
44.9667
45.1000
45.1333
45.2333
45.5667
46.1667
46.8333
47.5333
48.0333
48.5667
48.7667
49.1333
49.4333
49.8000
49.8667
49.8333
49.7000
50.0667
50.3333
50.7667
50.7000
50.8667
50.8000
50.7333
50.0667

‫اﻵن ﻧﺨﺘﺒﺮ ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬
MTB > TTest 0.0 'RESI1';
SUBC>
Alternative 0.

T-Test of the Mean

١٤٩

Test of mu = 0.000 vs mu not = 0.000
Variable
RESI1

N
57

Mean
0.158

StDev
0.868

SE Mean
0.115

T
1.37

P
0.17

‫ ﻳﺴﺘﺨﺪم ﺑﺮﻧﺎﻣﺞ ﻋﺎم ﻋﻨﺪﻣﺎ ﻳﻜﻮن اﻹﻧﺤﺮاف اﻟﻤﻌﻴﺎري ) او اﻟﺘﺒﺎﻳﻦ ( ﻏﻴﺮ‬Minitab ‫ ﻓﻲ‬:‫ﻣﻼﺣﻈﺔ‬
‫ أي‬١٫٩٦ ‫ وهﻲ اﻗﻞ ﻣﻦ‬T=١٫٣٧ ‫ ﻻﺣﻆ ان ﻗﻴﻤﺔ اﻹﺣﺼﺎﺋﺔ هﻲ‬. Ttest ‫ﻣﻌﺮوف وﻳﻄﻠﻖ ﻋﻠﻴﻪ‬
‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ‬
‫ ﺣﻮل اﻟﻤﺘﻮﺳﻂ وﺣﻮل اﻟﺼﻔﺮ‬Runs test ‫ ﻧﺨﺘﺒﺮ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ ﺑﻮاﺳﻄﺔ إﺧﺘﺒﺎر اﻟﺠﺮي‬:‫ﺛﺎﻧﻴﺎ‬
:‫ﺗﺎﺑﻊ اﻟﻤﺜﺎل‬
MTB > Runs 'RESI1'.

Runs Test

RESI1
K =

0.1579

The observed number of runs = 17
The expected number of runs = 29.4211
30 Observations above K
27 below
The test is significant at 0.0009
MTB > Runs 0 'RESI1'.

Runs Test

RESI1
K =

0.0000

The observed number of runs = 17
The expected number of runs = 28.7895
33 Observations above K
24 below
The test is significant at 0.0013

‫ﻧﻼﺣﻆ اﻧﻪ ﻓﻲ آﻠﺘﺎ اﻟﺤﺎﻟﺘﻴﻦ ﻻﻧﺮﻓﺾ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬
Autocorrelation test ‫ ﻧﺨﺘﺒﺮ ﺗﺮاﺑﻂ أو إﺳﺘﻘﻼل اﻟﺒﻮاﻗﻲ ﺑﻮاﺳﻄﺔ إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬:‫ﺛﺎﻟﺜﺎ‬
:‫ﺗﺎﺑﻊ اﻟﻤﺜﺎل‬
MTB > %ACF 'RESI1'.
Executing from file: E:\MTBWIN\MACROS\ACF.MAC

١٥٠

Autocorrelation

Autocorrelation Function for RESI1
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

4

9

14

Lag

Corr

T

LBQ

Lag

Corr

T

LBQ

1
2
3
4
5
6
7

0.56
0.24
-0.01
0.04
0.02
-0.07
-0.14

4.24
1.39
-0.06
0.21
0.11
-0.38
-0.81

18.93
22.31
22.32
22.40
22.42
22.72
24.07

8
9
10
11
12
13
14

-0.04
0.12
0.29
0.24
0.20
0.05
0.03

-0.22
0.68
1.64
1.30
1.05
0.25
0.17

24.17
25.21
31.36
35.67
38.71
38.90
38.99

‫ أي‬ρ1 = 0 ‫ أي اﻧﻨﺎ ﻧﺮﻓﺾ ان‬٤٫٢٤ ‫ ﻟﻠﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻋﻨﺪ اﻟﺘﺨﻠﻒ اﻷول ﺗﺴﺎوي‬T ‫ﻻﺣﻆ ان اﻟـ‬
‫ﻳﻮﺟﺪ ﺗﺮاﺑﻂ ﺑﻴﻦ اﻟﺒﻮاﻗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻻوﻟﻰ ﻓﻲ اﻹﺧﺘﺒﺎر‬
H 0 : ρ1 = 0

H1 : ρ1 ± 0
r1
= 4.24 ‫ﺣﻴﺚ اﻹﺣﺼﺎﺋﺔ هﻲ‬
se ( r1 )
‫ ﻧﺨﺘﺒﺮ ﻓﻲ ﻣﺎ إذا آﺎﻧﺖ اﻟﺒﻮاﻗﻲ ﻣﻮزﻋﺔ ﻃﺒﻴﻌﻴﺎ‬:‫راﺑﻌﺎ‬
:‫ﺗﺎﺑﻊ اﻟﻤﺜﺎل‬
MTB > %NormPlot 'RESI1';
SUBC>
Kstest.
Executing from file: E:\MTBWIN\MACROS\NormPlot.MAC

Normal Probability Plot

.999

P
ro
b
a
b
ility

.99
.95
.80
.50
.20
.05
.01
.001
-2

-1

0

1

RESI1
Average: 0.157895
StDev: 0.867525
N: 57

Kolmogorov-Smirnov Normality Test
D+: 0.054 D-: 0.084 D : 0.084
Approximate P-Value > 0.15

‫ أي ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ اﻟﺘﻮزﻳﻊ‬٠٫٠٥ ‫ وهﻲ اآﺒﺮ ﻣﻦ‬٠٫١٥ ‫ اﻟﻨﺎﺗﺠﺔ ﺗﺴﺎوي‬P-Value ‫ﻻﺣﻆ اﻟـ‬
‫ واﻟﺬي ﻳﺒﻴﻦ ﻣﺪي‬Q-Q Plot ‫ هﻨﺎك اﻳﻀﺎ إﺧﺘﺒﺎر ﺁﺧﺮ ﻟﻠﻄﺒﻴﻌﻴﺔ هﻮ اﻟـ‬α = 0.05 ‫اﻟﻄﺒﻴﻌﻲ ﻋﻨﺪ‬
‫ﺗﻄﺎﺑﻖ ﻣﺸﺎهﺪات ﻣﺎ ﻣﻊ ﺗﻮزﻳﻊ ﻣﻌﻴﻦ‬
MTB > %Qqplot 'RESI1';
SUBC>
Table;
SUBC>
Conf 95;
SUBC>
Ci.
Executing from file: E:\MTBWIN\MACROS\Qqplot.MAC

١٥١

‫‪Distribution Function Analysis‬‬

‫‪Normal Dist. Parameter Estimates‬‬
‫‪Data‬‬

‫‪: RESI1‬‬

‫‪Mean:‬‬
‫‪StDev:‬‬

‫‪0.157895‬‬
‫‪0.867525‬‬

‫‪Normal Probability Plot for RESI1‬‬

‫‪99‬‬

‫‪0.157895‬‬

‫‪Mean:‬‬

‫‪0.867525‬‬

‫‪StDev:‬‬
‫‪95‬‬
‫‪90‬‬
‫‪80‬‬

‫‪60‬‬
‫‪50‬‬
‫‪40‬‬

‫‪Percent‬‬

‫‪70‬‬

‫‪30‬‬
‫‪20‬‬
‫‪10‬‬
‫‪5‬‬

‫‪1‬‬

‫‪2‬‬

‫‪0‬‬

‫‪1‬‬

‫‪-1‬‬

‫‪-2‬‬

‫‪Data‬‬

‫ﻻﺣﻆ ان ﻓﻲ آﻠﺘﺎ اﻟﺤﺎﻟﺘﻴﻦ ﻓﺈﻧﻨﺎ ﻻﻧﺮﻓﺾ ان اﻟﺒﻮاﻗﻲ ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ‬
‫ﻣﻼﺣﻈﺔ اﺧﻴﺮة‪ :‬ﻳﺒﺪو ان اﻟﺒﻮاﻗﻲ ﺗﺤﻘﻖ ﻣﻌﻈﻢ اﻟﺸﺮوط ﻓﻴﻤﺎ ﻋﺪي اﻟﺘﺮاﺑﻂ اﻟﺬي ﻳﻮﺟﺪ ﺑﻴﻦ اﻟﻘﻴﻢ‬
‫اﻟﻤﺘﺘﺎﻟﻴﺔ وهﺬا ﻳﺠﻌﻠﻨﺎ ﻧﺮﻓﺾ ﺟﻮدة اﻟﺘﻄﺒﻴﻖ ﻟﻄﺮﻳﻘﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻟﺜﺔ ﺣﻴﺚ‬
‫ادي اﻟﻲ ﺑﻮاﻗﻲ ﻣﺘﺮاﺑﻄﺔ‪.‬‬

‫‪١٥٢‬‬

‫اﻟﻔﺼﻞ اﻟﺘﺎﺳﻊ‬
‫ﺗﺤﻠﻴﻞ او ﺗﻔﻜﻴﻚ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻲ ﻣﺮآﺒﺎت ‪:Decomposition Method‬‬
‫ﻳﻨﻈﺮ إﻟﻰ اﻟﻤﺘﻠﺴﻠﺴﺔ اﻟﺰﻣﻨﻴﺔ ﻋﻠﻰ اﻧﻬﺎ ﻣﻜﻮﻧﺔ ﻣﻦ ﻋﺪة ﻣﺮآﺒﺎت أو اﺟﺰاء ﻣﺘﺤﺪة ﻣﻊ ﺑﻌﻀﻬﺎ ﻟﺘﻜﻮﻳﻦ‬
‫هﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ‪،‬‬
‫ﻟﻨﻔﺘﺮض ان ﻟﺪﻳﻨﺎ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪﻩ ‪ . z1 , z2 ,..., zn‬ﻟﻘﺪ وﺟﺪ ﺑﺎﻟﺘﺠﺮﺑﺔ أﻧﻪ ﻳﻤﻜﻦ ﻧﻤﺬﺟﺘﻬﺎ‬
‫ﻋﻠﻰ اﻟﺸﻜﻞ‬
‫‪zt = Tt + St + Ct + Et , t = 1, 2,..., n‬‬

‫ﺣﻴﺚ ‪ Zt‬اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪة و ‪ Tt‬ﻣﺮآﺒﺔ اﻹﻧﺠﺮاف وهﻲ اﻟﺘﻲ ﺗﻨﻤﺬج اﻹﺗﺠﺎﻩ اﻟﻌﺎم اﻟﺬي‬
‫ﺗﻨﺤﻲ أو ﺗﻨﺠﺮف اﻟﻴﻪ اﻟﻤﺘﺴﻠﺴﻠﺔ و ‪ St‬ﻣﺮآﺒﺔ ﻣﻮﺳﻤﻴﺔ وﺗﻨﻤﺬج اﻟﺘﺄﺛﻴﺮ اﻟﻤﻮﺳﻤﻲ )إذا وﺟﺪ( وهﻮ‬
‫اﻟﺘﻐﻴﺮ اﻟﺬي ﻳﺤﺪث ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﻧﺘﻴﺠﺔ اﻟﺘﺄﺛﻴﺮات اﻟﻤﻮﺳﻤﻴﺔ ﻣﺜﻞ اﻟﺸﻬﺮﻳﺔ واﻟﺴﻨﻮﻳﺔ و ‪ Ct‬ﻣﺮآﺒﺔ‬
‫دورﻳﺔ )إذا وﺟﺪت( و ﺗﻨﻤﺬج ﻣﻨﺤﻰ أو إﺗﺠﺎة ﻳﺘﻜﺮر ﺑﻌﺪ ﻓﺘﺮات زﻣﻨﻴﺔ ﻃﻮﻳﻠﺔ ﻏﻴﺮ ﻣﻮﺳﻤﻴﺔ و ‪Et‬‬
‫ﻣﺮآﺒﺔ اﻟﺨﻄﺄ وﺗﺸﻤﻞ ﺟﻤﻴﻊ اﻟﻌﻮاﻣﻞ اﻻﺧﺮى اﻟﺘﻲ ﺗﺆﺛﺮ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ واﻟﺘﻲ ﻻﻳﻤﻜﻦ ﻧﻤﺬﺟﺘﻬﺎ‬
‫ﺿﻤﻨﻴﺎ أو اﻟﺘﻲ ﻻﻳﻤﻜﻦ ﻣﺸﺎهﺪﺗﻬﺎ او ﻗﻴﺎﺳﻬﺎ‪ .‬اﻟﻨﻤﻮذج اﻟﺴﺎﺑﻖ ﻳﺴﻤﻲ ﺑﺎﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ ‪Additive‬‬
‫‪ Model‬وذﻟﻚ ﻷن آﻞ اﻟﻤﺮآﺒﺎت ﺗﺪﺧﻞ ﺑﺸﻜﻞ إﺿﺎﻓﻲ ﻓﻲ اﻟﻨﻤﻮذج‪ .‬هﻨﺎك أﺷﻜﺎل اﺧﺮى ﻣﺜﻞ‬
‫‪zt = Tt St + Ct + Et , t = 1, 2,..., n‬‬
‫‪zt = Tt St Ct + Et , t = 1, 2,..., n‬‬

‫واﻟﺘﻲ ﺗﺴﻤﻰ ﺑﺎﻟﻨﻤﺎذج اﻟﺘﻀﺎﻋﻔﻴﺔ ‪. Multiplicative Models‬‬
‫ﻓﻲ هﺬا اﻟﻤﺴﺘﻮى ﺳﻮف ﻧﻬﻤﻞ اﻟﻤﺮآﺒﺔ اﻟﺪورﻳﺔ ‪ Ct‬وذﻟﻚ ﻷن اﻟﻤﺮآﺒﺔ اﻟﺪورﻳﺔ ﻧﺎدرا ﻣﺎﺗﻜﻮن‬
‫ﻣﻮﺟﻮدة ﻓﻲ اﻟﻤﺘﺴﻠﺴﻼت اﻟﻘﺼﻴﺮة أو اﻟﻄﻮﻳﻠﺔ ﻧﺴﺒﻴﺎ ﻷﻧﻬﺎ ﺗﺤﺘﺎج اﻟﻰ ﻣﺸﺎهﺪات ﻃﻮﻳﻠﺔ ﺟﺪا ﻋﻠﻰ‬
‫ﻣﺪي ﻋﺪد آﺒﻴﺮ ﻣﻦ اﻟﻌﻘﻮد‪.‬‬
‫وﻧﻜﺘﻔﻲ ﺑﺎﻟﻨﻤﺎذج ﻋﻠﻲ اﻟﺸﻜﻞ‬
‫‪zt = Tt + St + Et , t = 1,2,..., n‬‬

‫‪zt = Tt St + Et , t = 1,2,..., n‬‬
‫أﻧﻈﺮ آﺘﺎب ‪ FORECASTING: METHODS AND APPLICATIONS‬ﻟﻠﻤﺆﻟﻔﻴﻦ‬
‫‪ MAKRIDAKIS/ WHEELWRIGHT/ McGEE‬ص ‪١٤١-١٣١‬‬
‫ﺳﻮف ﻧﺴﺘﻌﺮض ﻓﻲ اﻟﻤﺜﺎل اﻟﺘﺎﻟﻲ ﻃﺮق اﻟﺘﺤﻠﻴﻞ اﻹﺿﺎﻓﻴﺔ واﻟﺘﻀﺎﻋﻔﻴﺔ ﺑﺪون اﻟﻤﺮآﺒﺔ اﻟﺪورﻳﺔ أي‬
‫اﻟﻨﻤﺎذج‬
‫‪zt = Tt + St + Et , t = 1,2,..., n‬‬

‫‪zt = Tt St + Et , t = 1,2,..., n‬‬
‫اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ هﻲ اﻟﻄﻠﺐ ﻋﻠﻲ اﻟﺒﻨﺰﻳﻦ ﺑﻤﻼﻳﻴﻦ اﻟﻠﺘﺮات ﻓﻲ ﻣﺪﻳﻨﺔ اوﻧﺘﺎرﻳﻮ ﺑﻜﻨﺪة ﻣﻦ ﺳﻨﺔ ‪١٩٦٠‬‬
‫وﺣﺘﻰ ﺳﻨﺔ ‪١٩٧٥‬‬
‫‪GasDemand‬‬
‫‪MONTHLY GASOLINE DEMAND ONTARIO GALLON MILLIONS 1960-1975‬‬
‫‪87695 86890 96442 98133 113615 123924 128924 134775 117357 114626‬‬
‫‪107677 108087 92188 88591 98683 99207 125485 124677 132543 140735‬‬
‫‪124008 121194 111634 111565 101007 94228 104255 106922 130621 125251‬‬
‫‪140318 146174 122318 128770 117518 115492 108497 100482 106140 118581‬‬
‫‪132371 132042 151938 150997 130931 137018 121271 123548 109894 106061‬‬
‫‪112539 125745 136251 140892 158390 148314 144148 140138 124075 136485‬‬
‫‪109895 109044 122499 124264 142296 150693 163331 165837 151731 142491‬‬
‫‪140229 140463 116963 118049 137869 127392 154166 160227 165869 173522‬‬
‫‪155828 153771 143963 143898 124046 121260 138870 129782 162312 167211‬‬
‫‪172897 189689 166496 160754 155582 145936 139625 137361 138963 155301‬‬
‫‪172026 165004 185861 190270 163903 174270 160272 165614 146182 137728‬‬
‫‪148932 156751 177998 174559 198079 189073 175702 180097 155202 174508‬‬

‫‪١٥٣‬‬

154277
174176
198688
218099
227443
193522
199024
217775

144998
184416
190474
229001
233038
212870
191813
227621

159644
158167
194502
203200
234119
248565
195997

168646
156261
190755
212557
255133
221532
208684

166273
176353
166286
197095
216478
252642
244113

190176
175720
170699
193693
232868
255007
243108

205541
193939
181468
188992
221616
206826
255918

193657
201269
174241
175347
209893
233231
244642

182617
218960
210802
196265
194784
212678
237579

189614
209861
212262
203526
189756
217173
237579

‫أوﻻ ﻧﺮﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﻲ ﻣﺨﻄﻂ زﻣﻨﻲ‬
MTB > TSPlot 'GasDemand';
SUBC>
Index;
SUBC>
TDisplay 11;
SUBC>
Symbol;
SUBC>
Connect.

G
asD
em
and

250

200

150

100

100
150
‫اﻟﻰ اﻻﻋﻠﻰ‬
‫ﻣﻮﺳﻤﻴﺔ وﻣﻨﺠﺮﻓﺔ‬
‫ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ‬
zt = Tt + St + Et , t = 1, 2,..., n :‫ ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ‬:‫اوﻻ‬
50

Index

SUBC>
SUBC>
SUBC>
SUBC>

MTB > %Decomp 'GasDemand' 12;
Additive ;
Forecasts 24;
Title "Forecast of Gasoline Demand";
Start 1.

Time Series Decomposition
Data
Length
NMissing

GasDeman
192.000
0

Trend Line Equation
Yt = 96.4074 + 0.680579*t
Seasonal Indices
Period
1
2
3
4
5
6
7
8
9
10
11
12

Index
-20.5625
-26.8125
-14.8958
-11.0625
9.89583
11.8958
22.7708
25.1875
5.64583
7.27083
-4.81250
-4.52083

١٥٤

Accuracy of Model
MAPE:
MAD:
MSD:

3.6952
5.6622
52.7851

Forecasts
Row

Period

Forecast

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216

207.197
201.627
214.225
218.738
240.377
243.058
254.614
257.711
238.850
241.155
229.753
230.725
215.364
209.794
222.391
226.905
248.544
251.225
262.780
265.878
247.017
249.322
237.919
238.892

Forecast of Gasoline Demand
Seasonal Indices

Original Data, by Seasonal Period

30

250

20
10

200

0
150

-10
-20

100

-30
1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12

Percent Variation, by Seasonal Period

Residuals, by Seasonal Period
30
20

10

10
0
5

-10
-20

0

-30
1 2 3 4 5 6 7 8 9 10 11 12

١٥٥

1 2 3 4 5 6 7 8 9 10 11 12

(١) ‫ﺷﻜﻞ‬

‫ﺷﻜﻞ )‪(٢‬‬
‫‪Forecast of Gasoline Demand‬‬
‫‪Original Data‬‬

‫‪Detrended Data‬‬
‫‪50‬‬
‫‪40‬‬
‫‪30‬‬
‫‪20‬‬
‫‪10‬‬
‫‪0‬‬
‫‪-10‬‬
‫‪-20‬‬
‫‪-30‬‬
‫‪-40‬‬
‫‪100‬‬

‫‪200‬‬

‫‪250‬‬
‫‪200‬‬
‫‪150‬‬
‫‪100‬‬
‫‪200‬‬

‫‪0‬‬

‫‪Seasonally Adj. and Detrended Data‬‬

‫‪0‬‬

‫‪100‬‬

‫‪Seasonally Adjusted Data‬‬
‫‪250‬‬

‫‪30‬‬
‫‪20‬‬
‫‪10‬‬

‫‪200‬‬

‫‪-10‬‬

‫‪150‬‬

‫‪0‬‬
‫‪-20‬‬
‫‪-30‬‬
‫‪100‬‬

‫‪200‬‬

‫ﺷﻜﻞ )‪(٣‬‬

‫‪100‬‬
‫‪200‬‬

‫‪0‬‬

‫‪0‬‬

‫‪100‬‬

‫‪Forecast of Gasoline Demand‬‬

‫‪Actual‬‬
‫‪Predicted‬‬

‫‪250‬‬

‫‪Actual‬‬
‫‪Predicted‬‬
‫‪Forecast‬‬

‫‪200‬‬

‫‪Forecast‬‬

‫‪GasDeman‬‬

‫‪150‬‬

‫‪100‬‬
‫‪3.6952‬‬
‫‪5.6622‬‬
‫‪52.7851‬‬

‫‪MAPE:‬‬
‫‪MAD:‬‬
‫‪MSD:‬‬

‫‪200‬‬

‫‪100‬‬

‫‪0‬‬

‫‪Time‬‬

‫ﻣﻨﺎﻗﺸﺔ اﻟﻨﺘﺎﺋﺞ‪:‬‬
‫ﺷﻜﻞ )‪ (١‬ﻳﻮﺿﺢ اﻟﻤﺆﺷﺮات اﻟﻤﻮﺳﻤﻴﺔ ‪ ، Seasonal Indices‬ﻓﺎﻟﺸﻜﻞ اﻷﻋﻠﻰ ﻣﻦ اﻟﻴﺴﺎر ﻳﺒﻴﻦ‬
‫ﺗﺄﺛﺮاﻟﻄﻠﺐ ﻓﻲ اﻷﺷﻬﺮ اﻟﻤﺨﺘﻠﻔﺔ ﻣﻦ اﻟﺴﻨﺔ ﻓﻔﻲ اﻷﺷﻬﺮ ‪ ١١‬و ‪ ١٢‬و ‪ ١‬و ‪ ٢‬و ‪ ٣‬و ‪ ٤‬ﻳﺤﺪث ﻧﻘﺺ‬
‫ﻓﻲ اﻟﻄﻠﺐ إذ ﻳﺘﻨﺎﻗﺺ ﺗﺪرﻳﺠﻴﺎ ﺣﺘﻰ ﻳﺼﻞ إﻟﻰ أﻗﻞ ﻣﻌﺪل ﻟﻪ ﻓﻲ اﻟﺸﻬﺮ ‪ ٢‬ﺛﻢ ﻳﺘﺰاﻳﺪ ﺣﺘﻰ ﻳﺼﺒﺢ‬
‫ﻣﻮﺟﺒﺎ ﻓﻲ اﻟﺸﻬﺮ ‪ ٥‬وﻳﺘﺰاﻳﺪ ﺣﺘﻰ ﻳﺼﻞ اﻗﺼﻰ ﻗﻴﻤﺔ ﻣﻮﺟﺒﺔ ﻓﻲ اﻟﺸﻬﺮ ‪ ٨‬ﺛﻢ ﻳﻨﻘﺺ ﺑﺸﻜﻞ آﺒﻴﺮ‬
‫ﺑﻌﺪﺋﺬ‪ .‬اﻟﺸﻜﻞ اﻷﻋﻠﻰ ﻣﻦ اﻟﻴﻤﻴﻦ ﻳﻌﻄﻲ رﺳﻢ اﻟﺼﻨﺪوق ‪ Box Plot‬ﻟﻠﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ ﻣﻮزﻋﺔ‬
‫ﻋﻠﻰ اﻷﺷﻬﺮ وهﻮ ﻳﻮﺿﺢ ﺗﻮزﻳﻊ وإﻧﺘﺸﺎر اﻟﻤﺸﺎهﺪات ﻋﻠﻰ آﻞ ﺷﻬﺮ واﻟﻘﻴﻢ اﻟﺨﺎرﺟﺔ ‪. Out Liers‬‬
‫اﻟﺸﻜﻞ اﻷﺳﻔﻞ ﻣﻦ اﻟﻴﺴﺎر ﻳﻌﻄﻲ اﻟﺘﻐﻴﺮ اﻟﻨﺴﺒﻲ اﻟﻤﺌﻮي ﻋﻠﻰ اﻟﻔﺘﺮات اﻟﻤﻮﺳﻤﻴﺔ )اﻷﺷﻬﺮ(‪ .‬اﻟﺸﻜﻞ‬
‫اﻷﺳﻔﻞ اﻷﻳﻤﻦ ﻳﻌﻄﻲ رﺳﻢ اﻟﺼﻨﺪوق ﻟﻠﺒﻮاﻗﻲ أو اﻷﺧﻄﺎء ﻣﻮزﻋﺔ ﻋﻠﻰ اﻷﺷﻬﺮ‪.‬‬
‫ﺷﻜﻞ )‪ (٢‬اﻟﺸﻜﻞ اﻷﻋﻠﻰ ﻣﻦ اﻟﻴﻤﻴﻦ ﻳﻌﻄﻲ اﻟﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ‪ ،‬اﻟﺸﻜﻞ اﻷﻋﻠﻰ ﻣﻦ اﻟﻴﺴﺎر ﻳﻌﻄﻲ‬
‫اﻟﻤﺸﺎهﺪات ﺑﻌﺪ إزاﺣﺔ ﻣﺮآﺒﺔ اﻹﻧﺠﺮاف أي‬
‫‪wt = zt − Tt , t = 1, 2,..., n‬‬
‫‪=St + Et , t = 1, 2,..., n‬‬

‫‪١٥٦‬‬

‫اﻟﺸﻜﻞ اﻷﺳﻔﻞ ﻣﻦ اﻟﻴﺴﺎر ﻳﻌﻄﻲ اﻟﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ ﺑﻌﺪ إزاﺣﺔ اﻟﻤﺮآﺒﺔ اﻟﻤﻮﺳﻤﻴﺔ أي‬
yt = zt − St , t = 1, 2,..., n
=Tt + Et , t = 1, 2,..., n

‫ أو اﻟﺒﻮاﻗﻲ ﺑﻌﺪ إزاﺣﺔ ﻣﺮآﺒﺘﻲ اﻹﻧﺠﺮاف‬Et ‫اﻟﺸﻜﻞ اﻷﺳﻔﻞ اﻷﻳﻤﻦ ﻳﻌﻄﻲ ﻣﺮآﺒﺔ اﻟﺨﻄﺄ‬
‫واﻟﻤﻮﺳﻤﻴﺔ ﻣﻦ اﻟﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ أي‬
et = zt − Tt − St , t = 1, 2,..., n
=Et , t = 1, 2,..., n

.‫ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻣﻊ ﻣﻘﺎﻳﻴﺲ دﻗﺔ اﻟﺘﻄﺒﻴﻖ‬٢٤ ‫( ﻳﻌﻄﻲ اﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ‬٣) ‫ﺷﻜﻞ‬
zt = Tt St + Et , t = 1, 2,..., n

:‫ ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج اﻟﺘﻀﺎﻋﻔﻲ‬:‫ﺛﺎﻧﻴﺎ‬

MTB > %Decomp 'GasDemand' 12;
SUBC>
Forecasts 24;
SUBC>
Title "Forecast of Gasoline Demand";
SUBC> Start 1.
Executing from file: D:\MTBWIN\MACROS\Decomp.MAC
Macro is running ... please wait

Time Series Decomposition
Data
Length
NMissing

GasDeman
192.000
0

Trend Line Equation
Yt = 96.4074 + 0.680579*t
Seasonal Indices
Period

Index

1
2
3
4
5
6
7
8
9
10
11
12

0.860355
0.828555
0.892431
0.936273
1.06124
1.07274
1.15775
1.17075
1.03409
1.05059
0.966300
0.968923

Accuracy of Model
MAPE:
MAD:
MSD:

3.6338
5.7720
56.8996

Forecasts
Row

Period

Forecast

1
2

193
194

195.954
189.275

١٥٧

3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216

204.474
215.156
244.596
247.977
268.415
272.227
241.154
245.718
226.660
227.935
202.980
196.042
211.762
222.803
253.263
256.738
277.870
281.789
249.599
254.298
234.552
235.848

(٤) ‫ﺷﻜﻞ‬
Forecast of Gasoline Demand
Seasonal Indices

Original Data, by Seasonal Period

1.2

250

1.1

200

1.0
150
0.9
100
0.8
1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12

Percent Variation, by Seasonal Period
14
12
10
8
6
4
2
0

Residuals, by Seasonal Period
20
10
0
-10
-20
-30

1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12

(٥) ‫ﺷﻜﻞ‬
Forecast of Gasoline Demand
Original Data

Detrended Data

250

1.3
1.2

200

1.1
1.0

150

0.9
100

0.8
0

100

200

0

Seasonally Adjusted Data
240
220
200
180
160
140

١٥٨

100

200

Seasonally Adj. and Detrended Data
20
10
0
-10

‫‪Forecast of Gasoline Demand‬‬

‫ﺷﻜﻞ )‪(٦‬‬
‫‪Actual‬‬

‫‪280‬‬

‫‪Predicted‬‬
‫‪Forecast‬‬

‫‪180‬‬

‫‪3.6338‬‬
‫‪5.7720‬‬
‫‪56.8996‬‬

‫‪MAPE:‬‬
‫‪MAD:‬‬
‫‪MSD:‬‬

‫‪GasDeman‬‬

‫‪Actual‬‬
‫‪Predicted‬‬
‫‪Forecast‬‬

‫‪80‬‬

‫‪200‬‬

‫‪100‬‬

‫‪0‬‬

‫‪Time‬‬

‫ﻣﻨﺎﻗﺸﺔ اﻟﻨﺘﺎﺋﺞ‪:‬‬
‫اﻷﺷﻜﺎل )‪ (٤‬و )‪ (٥‬و )‪ (٦‬ﻟﻬﺎ ﻧﻔﺲ اﻟﺘﻔﺴﻴﺮ آﻤﺎ ﻓﻲ اﻷﺷﻜﺎل )‪ (١‬و )‪ (٢‬و )‪.(٣‬‬
‫ﺑﻤﺎ اﻧﻨﺎ ﻃﺒﻘﻨﺎ ﻧﻤﻮذﺟﻴﻦ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺸﺎهﺪة ﻓﻴﺠﺐ أن ﻧﺨﺘﺎر اﻓﻀﻞ ﻧﻤﻮذج‪ ،‬وهﻨﺎ ﻳﺄﺗﻲ دور‬
‫ﻣﻘﺎﻳﻴﺲ دﻗﺔ اﻟﺘﻄﺒﻴﻖ واﻟﺘﻲ ﺗﻨﺘﺞ ﻣﻦ اﻟﺒﺮﻧﺎﻣﺞ‪ ،‬ﻟﺪﻳﻨﺎ ﺛﻼﺛﺔ ﻣﻘﺎﻳﻴﺲ دﻗﺔ‪:‬‬
‫‪ -١‬ﻣﺘﻮﺳﻂ اﻟﺨﻄﺄ اﻟﻨﺴﺒﻲ اﻟﻤﻄﻠﻖ ‪ Mean Absolute Percentage Error‬أو ‪MAPE‬‬
‫وﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬
‫‪zt − zˆt‬‬
‫‪zt‬‬
‫‪× 100, zt ≠ 0‬‬
‫‪n‬‬

‫‪n‬‬

‫∑‬
‫‪t =1‬‬

‫= ‪MAPE‬‬

‫‪ -٢‬ﻣﺘﻮﺳﻂ اﻹﻧﺤﺮاف اﻟﻤﻄﻠﻖ ‪ Mean Absolute Deviation‬أو ‪ MAD‬وﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬
‫‪n‬‬

‫‪− zˆt‬‬

‫‪t‬‬

‫‪∑z‬‬
‫‪t =1‬‬

‫‪n‬‬

‫= ‪MAD‬‬

‫‪ -٣‬ﻣﺘﻮﺳﻂ أﻹﻧﺤﺮاف اﻟﻤﺮﺑﻊ )أو ﻣﺘﻮﺳﻂ اﻟﺨﻄﺄ اﻟﻤﺮﺑﻊ( ‪) MSD‬أو ‪ ( MSE‬وﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬
‫‪2‬‬

‫) ‪− zˆt‬‬

‫‪n‬‬

‫‪t‬‬

‫‪∑( z‬‬

‫‪n‬‬

‫‪t =1‬‬

‫= ‪MSD‬‬

‫ﺑﺈﺧﺘﻴﺎر أﺣﺪ هﺬة اﻟﻤﻘﺎﻳﻴﺲ ﻧﺨﺘﺎر اﻟﻨﻤﻮذج اﻟﺬي ﻳﻌﻄﻲ أﻗﻞ ﻗﻴﻤﺔ ﻟﻬﺬا اﻟﻤﻘﻴﺎس‪ ،‬اﻟﻤﻘﻴﺎس اﻷآﺜﺮ‬
‫إﺳﺘﺨﺪاﻣﺎ وﺷﻴﻮﻋﺎ هﻮ ‪ MSD‬أو ‪ MSE‬وهﻮ اﻟﺬي ﺳﻮف هﻨﺎ‪.‬‬
‫ﻟﻠﻨﻤﻮذج اﻹﺿﺎﻓﻲ ﻣﻘﺎﻳﻴﺲ اﻟﺪﻗﺔ هﻲ‪:‬‬
‫‪3.6952‬‬
‫‪١٥٩‬‬

‫‪MAPE:‬‬

‫‪MAD:‬‬
‫‪MSD:‬‬

‫‪5.6622‬‬
‫‪52.7851‬‬
‫و ﻟﻠﻨﻤﻮذج اﻟﺘﻀﺎﻋﻔﻲ‪:‬‬

‫‪MAPE:‬‬
‫‪3.6338‬‬
‫‪MAD:‬‬
‫‪5.7720‬‬
‫‪MSD:‬‬
‫‪56.8996‬‬
‫ﻧﻼﺣﻆ أن اﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ اﻋﻄﻰ اﻗﻞ ﻗﻴﻤﺔ ﻟﻠﻤﻘﻴﺎس ‪ MSD‬وﻟﺬﻟﻚ ﻧﻘﺮر إﺳﺘﺨﺪام هﺬا اﻟﻨﻤﻮذج‬
‫ﻟﻠﺘﻨﺒﺆ ﻋﻦ اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻄﻠﺐ‪.‬‬

‫ﺗﻮﺿﻴﺢ ﻃﺮﻳﻘﺔ ﺗﺤﻠﻴﻞ او ﺗﻔﻜﻴﻚ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻲ ﻣﺮآﺒﺎت ‪Decomposition‬‬
‫‪:Method‬‬
‫ﺳﻮف ﻧﻮﺿﺢ ﺑﺎﻟﻤﺜﺎل اﻟﺘﺎﻟﻲ ﻃﺮﻳﻘﺔ ﺗﺤﻠﻴﻞ أو ﺗﻔﻜﻴﻚ اﻟﻤﺘﺴﻠﺴﻠﺔ إﻟﻰ ﻣﺮآﺒﺎت‪ .‬اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ إﻧﺘﺎج‬
‫‪ ١٦٨‬ﻳﻮﻣﺎ ﻟﻠﺤﻠﻴﺐ ﻓﻲ أﺣﺪ اﻟﻤﺰارع ﺑﺎﻟﻜﻴﻠﻮ ﺟﺮام‬
‫‪MTB > Read "E:\Mtbwin\milk.dat" c1.‬‬
‫‪Entering data from file: E:\Mtbwin\milk.dat‬‬
‫‪168 rows read.‬‬
‫'‪MTB > name c1='MilkProd‬‬
‫‪MTB > print c1‬‬

‫‪Data Display‬‬
‫‪MilkProd‬‬
‫‪582‬‬
‫‪598‬‬
‫‪634‬‬
‫‪635‬‬
‫‪688‬‬
‫‪698‬‬
‫‪711‬‬
‫‪734‬‬
‫‪751‬‬
‫‪800‬‬
‫‪805‬‬
‫‪802‬‬
‫‪813‬‬
‫‪843‬‬

‫‪553‬‬
‫‪565‬‬
‫‪594‬‬
‫‪602‬‬
‫‪645‬‬
‫‪660‬‬
‫‪677‬‬
‫‪690‬‬
‫‪711‬‬
‫‪763‬‬
‫‪766‬‬
‫‪760‬‬
‫‪773‬‬
‫‪797‬‬

‫‪577‬‬
‫‪587‬‬
‫‪611‬‬
‫‪621‬‬
‫‪667‬‬
‫‪687‬‬
‫‪706‬‬
‫‪723‬‬
‫‪747‬‬
‫‪800‬‬
‫‪810‬‬
‫‪791‬‬
‫‪812‬‬
‫‪827‬‬

‫‪568‬‬
‫‪583‬‬
‫‪604‬‬
‫‪615‬‬
‫‪661‬‬
‫‪681‬‬
‫‪701‬‬
‫‪725‬‬
‫‪740‬‬
‫‪790‬‬
‫‪809‬‬
‫‪784‬‬
‫‪815‬‬
‫‪817‬‬

‫‪599‬‬
‫‪617‬‬
‫‪639‬‬
‫‪653‬‬
‫‪697‬‬
‫‪722‬‬
‫‪740‬‬
‫‪764‬‬
‫‪783‬‬
‫‪834‬‬
‫‪855‬‬
‫‪837‬‬
‫‪867‬‬
‫‪858‬‬

‫‪640‬‬
‫‪660‬‬
‫‪678‬‬
‫‪702‬‬
‫‪735‬‬
‫‪767‬‬
‫‪783‬‬
‫‪801‬‬
‫‪819‬‬
‫‪869‬‬
‫‪894‬‬
‫‪881‬‬
‫‪908‬‬
‫‪896‬‬

‫‪697‬‬
‫‪716‬‬
‫‪736‬‬
‫‪756‬‬
‫‪798‬‬
‫‪817‬‬
‫‪826‬‬
‫‪845‬‬
‫‪859‬‬
‫‪913‬‬
‫‪935‬‬
‫‪924‬‬
‫‪947‬‬
‫‪937‬‬

‫‪727‬‬
‫‪742‬‬
‫‪770‬‬
‫‪782‬‬
‫‪811‬‬
‫‪837‬‬
‫‪858‬‬
‫‪871‬‬
‫‪886‬‬
‫‪942‬‬
‫‪961‬‬
‫‪957‬‬
‫‪969‬‬
‫‪966‬‬

‫‪640‬‬
‫‪653‬‬
‫‪688‬‬
‫‪709‬‬
‫‪736‬‬
‫‪762‬‬
‫‪775‬‬
‫‪785‬‬
‫‪807‬‬
‫‪860‬‬
‫‪890‬‬
‫‪883‬‬
‫‪889‬‬
‫‪892‬‬

‫‪656‬‬
‫‪673‬‬
‫‪705‬‬
‫‪722‬‬
‫‪755‬‬
‫‪784‬‬
‫‪796‬‬
‫‪805‬‬
‫‪824‬‬
‫‪878‬‬
‫‪900‬‬
‫‪898‬‬
‫‪902‬‬
‫‪903‬‬

‫‪561‬‬
‫‪566‬‬
‫‪618‬‬
‫‪622‬‬
‫‪635‬‬
‫‪667‬‬
‫‪696‬‬
‫‪690‬‬
‫‪707‬‬
‫‪756‬‬
‫‪799‬‬
‫‪773‬‬
‫‪778‬‬
‫‪782‬‬

‫‪589‬‬
‫‪600‬‬
‫‪628‬‬
‫‪658‬‬
‫‪677‬‬
‫‪713‬‬
‫‪717‬‬
‫‪734‬‬
‫‪750‬‬
‫‪804‬‬
‫‪826‬‬
‫‪821‬‬
‫‪828‬‬
‫‪834‬‬

‫وﻧﺮﺳﻢ اﻟﺒﻴﺎﻧﺎت اﻟﺴﺎﺑﻘﺔ ﺑﺎﻷواﻣﺮ‪:‬‬
‫;'‪MTB > TSPlot 'MilkProd‬‬
‫>‪SUBC‬‬
‫;‪Index‬‬
‫>‪SUBC‬‬
‫;‪TDisplay 11‬‬
‫>‪SUBC‬‬
‫;‪Symbol‬‬
‫>‪SUBC‬‬
‫‪Connect.‬‬

‫‪1000‬‬

‫‪900‬‬

‫‪700‬‬

‫‪600‬‬

‫‪150‬‬

‫‪100‬‬

‫‪١٦٠‬‬

‫‪50‬‬

‫‪Index‬‬

‫‪MilkProd‬‬

‫‪800‬‬

:‫ ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ‬:‫اوﻻ‬
‫ ﻟﻜﻲ‬zt = Tt + St + Et , t = 1, 2,..., n ‫ ﻧﻜﺘﺐ اﻟﻨﻤﻮذج ﻋﻠﻰ اﻟﺸﻜﻞ‬z1 , z2 ,..., zn ‫ﻟﻤﺸﺎهﺪات‬
:‫ﻧﻘﻮم ﺑﺘﻔﻜﻴﻚ هﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻲ اﻟﻤﺮآﺒﺎت اﻟﺴﺎﺑﻘﺔ ﺳﻮف ﻧﺴﺘﻌﺮض ﻃﺮﻳﻘﺘﻴﻦ ﻣﻤﻜﻨﺔ‬
:‫اﻟﻄﺮﻳﻘﺔ اﻻوﻟﻲ‬
:‫ أي‬Tt ‫ ﻧﻄﺒﻖ إﻧﺤﺪار ﺧﻄﻲ ﺑﺴﻴﻂ ﻟﻠﻤﺸﺎهﺪات ﻋﻠﻲ اﻟﺰﻣﻦ ﻟﺘﻘﺪﻳﺮ ﻣﺮآﺒﺔ اﻻﻧﺠﺮاف‬-١
Tˆt ≡ zˆt = a + bt , t = 1, 2,...,168

:‫أي‬
MTB > set c2
DATA> 1:168
DATA> end
MTB > name c1='MilkProd' c2='Time' c3='Trend' c5='Detrend' c6='Index'
c8='Fitted' c9='Resid'
MTB > regr c1 1 c2;
SUBC> fits c3.

Regression Analysis
The regression equation is
MilkProd = 612 + 1.69 Time
Predictor
Constant
Time
S = 60.74

Coef
StDev
T
P
611.682
9.414
64.97
0.000
1.69262
0.09663
17.52
0.000
R-Sq = 64.9%
R-Sq(adj) = 64.7%

Analysis of Variance
Source
Regression
Error
Total

DF
1
166
167

SS
1132003
612439
1744443

MS
1132003
3689

F
306.83

P
0.000

‫وﺷﻜﻞ اﻹﻧﺠﺮاف هﻮ‬

900

Trend

800

700

600

Index

50

100

150

‫ ﻧﻄﺮح ﻣﺮآﺒﺔ اﻹﻧﺠﺮاف ﻣﻦ اﻟﻤﺸﺎهﺪات اﻻﺻﻠﻴﺔ ﻓﻨﺤﺼﻞ ﻋﻠﻲ ﻣﺎﻳﺴﻤﻲ ﺑﺎﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺰاﻟﺔ‬-٢
zt − zˆt = zt − Tˆt , t = 1, 2,...,168 ‫ أي‬Detrended Series ‫اﻹﻧﺠﺮاف‬
MTB > let c5=c1-c3

:‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬

١٦١

Detrend

100

0

-100
Index

50

100

150

zt − Tˆt = St + Et , t = 1, 2,...,168 ‫ﻧﻼﺣﻆ اﻧﻬﺎ اﺻﺒﺤﺖ ﻣﻮﺳﻤﻴﺔ ﻓﻘﻂ ﻷن‬
:‫ آﺎﻟﺘﺎﻟﻲ‬Seasonal Indices ‫ ﻟﺘﻘﺪﻳﺮ اﻟﻤﺮآﺒﺔ اﻟﻤﻮﺳﻤﻴﺔ ﻧﻮﺟﺪ اﻟﻤﺆﺷﺮات اﻟﻤﻮﺳﻤﻴﺔ‬-٣
‫ اﻟﻤﺆﺷﺮ اﻟﻤﻮﺳﻤﻲ ﻟﻠﺸﻬﺮ‬I1 ‫ ﺣﻴﺚ‬I s , s = 1, 2,...,12 ‫ﻟﻨﺮﻣﺰ ﻟﻠﻤﺆﺷﺮات اﻟﻤﻮﺳﻤﻴﺔ ﺑﺎﻟﺮﻣﺰ‬
d t = zt − Tˆt , t = 1, 2,...,168 ‫ اﻟﻤﺆﺷﺮ اﻟﻤﻮﺳﻤﻲ ﻟﻠﺸﻬﺮ اﻟﺜﺎﻧﻲ وهﻜﺬا وﻟﻨﺮﻣﺰ ﺑـ‬I 2 ‫اﻷول و‬
:‫ﺗﻘﺪر هﺬﻩ اﻟﻤﺆﺷﺮات آﺎﻟﺘﺎﻟﻲ‬

1
( d1 + d13 + d 25 + L + d157 )
14
1
I 2 = ( d 2 + d14 + d 26 + L + d158 )
14
M
I1 =

I12 =

1
( d12 + d 24 + d 36 + L + d168 )
14
:‫وﻳﺘﻢ ذﻟﻚ ﺑﺈﺳﺘﺨﺪام اﻷواﻣﺮ اﻟﺘﺎﻟﻴﺔ‬

MTB >
DATA>
DATA>
MTB >
SUBC>
SUBC>
MTB >
&
CONT>
MTB >
MTB >
MTB >
DATA>
DATA>
MTB >

set c4
14(1:12)
end
stat c5;
by c4;
mean c6.
Stack 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index'
'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' c7.
let c8=c3+c7
let c9=c1-c8
set c10
1:12
end
print c10 c6

Data Display

١٦٢

‫‪ -٤‬اﻟﺘﻨﺒﺆات ﺗﻮﻟﺪ آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫‪Index‬‬

‫‪Season‬‬

‫‪Row‬‬

‫‪-18.328‬‬
‫‪-57.806‬‬
‫‪34.716‬‬
‫‪49.595‬‬
‫‪110.616‬‬
‫‪82.281‬‬
‫‪32.517‬‬
‫‪-9.747‬‬
‫‪-52.297‬‬
‫‪-48.775‬‬
‫‪-79.754‬‬
‫‪-43.018‬‬

‫‪1‬‬
‫‪2‬‬
‫‪3‬‬
‫‪4‬‬
‫‪5‬‬
‫‪6‬‬
‫‪7‬‬
‫‪8‬‬
‫‪9‬‬
‫‪10‬‬
‫‪11‬‬
‫‪12‬‬

‫‪1‬‬
‫‪2‬‬
‫‪3‬‬
‫‪4‬‬
‫‪5‬‬
‫‪6‬‬
‫‪7‬‬
‫‪8‬‬
‫‪9‬‬
‫‪10‬‬
‫‪11‬‬
‫‪12‬‬

‫‪z168 ( l ) = 612 + 1.69 ( l + 168) + I ( l mod 12 ) , l = 1,2,...‬‬

‫ﻓﻤﺜﻼ اﻟﺘﻨﺒﺆ ﻋﻨﺪ اﻟﻴﻮم ‪ ١٦٩‬هﻮ‬

‫‪z168 (1) = 612 + 1.69 (169 ) + I1‬‬
‫‪=897.61 + ( −18.328 ) = 879.282‬‬

‫اﻟﻄﺮﻳﻘﺔ اﻟﺜﺎﻧﻴﺔ‪:‬‬
‫وهﻲ اﻟﺘﻲ ﻳﺴﺘﺨﺪﻣﻬﺎ ﺑﺮﻧﺎﻣﺞ ‪: Minitab‬‬
‫‪ -١‬آﺎﻟﻄﺮﻳﻘﺔ اﻻوﻟﻰ ﻧﻄﺒﻖ إﻧﺤﺪار ﺧﻄﻲ ﺑﺴﻴﻂ ﻟﻠﻤﺸﺎهﺪات ﻋﻠﻲ اﻟﺰﻣﻦ ﻟﺘﻘﺪﻳﺮ ﻣﺮآﺒﺔ اﻻﻧﺠﺮاف‬
‫‪ Tt‬ﻓﻨﺤﺼﻞ ﻋﻠﻰ ﻧﻔﺲ اﻟﻨﺘﻴﺠﺔ آﻤﺎ ﻓﻲ اﻟﻄﺮﻳﻘﺔ اﻻوﻟﻰ )‪(١‬‬
‫‪ -٢‬اﻳﻀﺎ هﻨﺎ ﻧﻄﺮح ﻣﺮآﺒﺔ اﻹﻧﺠﺮاف ﻣﻦ اﻟﻤﺸﺎهﺪات اﻻﺻﻠﻴﺔ ﻓﻨﺤﺼﻞ ﻋﻠﻲ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺰاﻟﺔ‬
‫اﻹﻧﺠﺮاف ‪Detrended Series‬‬
‫‪ -٣‬ﻧﻄﺒﻖ اﻵن ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ درﺟﺔ اﻟﻤﻮﺳﻢ وﻧﻮﺳﻄﻪ اذا اﺣﺘﺎج اﻻﻣﺮ‬
‫‪ -٤‬ﻧﻄﺮح اﻟﻤﺘﻮﺳﻄﺎت اﻟﻤﺘﺤﺮآﺔ ﻣﻦ ﻧﻈﻴﺮاﺗﻬﺎ ﻓﻲ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺰاﻟﺔ اﻹﻧﺠﺮاف ﻓﻨﺤﺼﻞ ﻋﻠﻲ‬
‫ﻣﺘﺴﻠﺴﻠﺔ ﺗﺤﻮي اﻟﻤﺮآﺒﺎت اﻟﻤﻮﺳﻤﻴﺔ‬
‫‪ -٥‬ﺗﻘﺪر اﻟﻤﺮآﺒﺎت اﻟﻤﻮﺳﻤﻴﺔ آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫) ‪I1 = Median ( d1 , d13 , d 25 ,L, d157‬‬

‫) ‪I 2 = Median ( d 2 , d14 , d 26 ,L, d158‬‬

‫‪M‬‬

‫) ‪I12 = Median ( d12 , d 24 , d 36 ,L, d168‬‬
‫‪ -٦‬ﺗﻮﻟﺪ اﻟﺘﻨﺒﺆات آﺎﻟﺴﺎﺑﻖ‬
‫وﺳﻮف ﻧﺴﺘﻌﺮض هﺬا آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫‪MTB > Read "E:\Mtbwin\milk.dat" c1.‬‬
‫‪Entering data from file: E:\Mtbwin\milk.dat‬‬
‫‪168 rows read.‬‬
‫'‪MTB > name c1='MilkProd‬‬
‫‪MTB > set c2‬‬
‫‪DATA> 1:168‬‬
‫‪DATA> end‬‬
‫'‪MTB > name c2='Time‬‬
‫;‪MTB > regr c1 1 c2‬‬
‫‪SUBC> fits c3.‬‬

‫‪Regression Analysis‬‬

‫‪١٦٣‬‬

The regression equation is
MilkProd = 612 + 1.69 Time
Predictor
Constant
Time

Coef
611.682
1.69262

S = 60.74

StDev
9.414
0.09663

R-Sq = 64.9%

T
64.97
17.52

P
0.000
0.000

R-Sq(adj) = 64.7%

Analysis of Variance
Source
Regression
Error
Total

DF
1
166
167

SS
1132003
612439
1744443

Unusual Observations
Obs
Time
MilkProd
113
113
942.00
125
125
961.00

MS
1132003
3689

Fit
802.95
823.26

F
306.83

StDev Fit
5.44
6.11

P
0.000

Residual
139.05
137.74

St Resid
2.30R
2.28R

R denotes an observation with a large standardized residual
MTB
MTB
MTB
MTB

>
>
>
>

name c3='Trend'
let c4=c1-c3
name c4='Detrend'
Name c5 = 'AVER1'

:‫ وﻧﻮﺳﻄﻪ‬١٢ ‫ﻓﻲ اﻟﺨﻄﻮة اﻟﺘﺎﻟﻴﺔ ﻧﻄﺒﻖ ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ‬
MTB > %MA 'Detrend' 12;
SUBC> Center;
SUBC>
Averages 'AVER1'.
Executing from file: E:\MTBWIN\MACROS\MA.MAC
Macro is running ... please wait

Moving average
Data
Length
NMissing

Detrend
168.000
0

Moving Average
Length: 12
Accuracy Measures
MAPE: 111.68
MAD:
52.36
MSD: 3564.77

‫ﻧﻄﺮح اﻟﻤﺘﻮﺳﻄﺎت اﻟﻤﺘﺤﺮآﺔ اﻟﻤﻮﺳﻄﺔ ﻣﻦ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺰال إﻧﺠﺮاﻓﻬﺎ‬
MTB >
MTB >
MTB >
DATA>
DATA>
MTB >
SUBC>
SUBC>
MTB >

let c6=c4-c5
name c6='DeSeason'
set c2
14(1:12)
end
stat c6;
by c2;
median c7.
name c7='SeasInx'

Data Display
Row

Season

SeasInx

١٦٤

1
2
3
4
5
6
7
8
9
10
11
12

1
2
3
4
5
6
7
8
9
10
11
12

-20.750
-58.958
35.625
50.083
109.542
81.292
33.917
-10.000
-52.792
-50.250
-79.958
-44.375

:‫ﻧﻘﺎرن اﻟﺤﺴﺎﺑﺎت اﻟﺘﻲ اﺟﺮﻳﻨﺎهﺎ ﻣﻊ اﻟﺒﺮﻧﺎﻣﺞ اﻷﺻﻠﻲ‬
MTB > %Decomp 'MilkProd' 12;
SUBC> Additive ;
SUBC> Start 1.
Executing from file: E:\MTBWIN\MACROS\Decomp.MAC
Macro is running ... please wait

Time Series Decomposition
Data
Length
NMissing

MilkProd
168.000
0

Trend Line Equation
Yt = 611.682 + 1.69262*t
Seasonal Indices
Period
1
2
3
4
5
6
7
8
9
10
11
12

Index
-20.1979
-58.4062
36.1771
50.6354
110.094
81.8437
34.4687
-9.44792
-52.2396
-49.6979
-79.4063
-43.8229

Accuracy of Model
MAPE:
MAD:
MSD:

1.583
12.088
244.406

.‫وﺑﻤﻘﺎرﻧﺔ اﻟﻨﺘﻴﺠﺘﻴﻦ ﻧﺠﺪ اﻧﻬﻤﺎ ﺗﻘﺮﻳﺒﺎ ﻣﺘﺴﺎوﻳﺘﺎن‬
:‫ آﺎﻟﺘﺎﻟﻲ‬%Decomp ‫ﺳﻮف ﻧﻮﻟﺪ ﺗﻨﺒﺆات ﺑﺎﺳﺘﺨﺪام اﻟﺒﺮﻧﺎﻣﺞ‬
MTB > %Decomp 'MilkProd' 12;
SUBC> Additive ;
SUBC>
Forecasts 12;
SUBC> Start 1.

:‫واﻟﺘﻲ ﺗﻌﻄﻲ اﻟﺘﻨﺒﺆات‬
Forecasts

١٦٥

‫‪Forecast‬‬

‫‪Period‬‬

‫‪Row‬‬

‫‪877.54‬‬
‫‪841.02‬‬
‫‪937.30‬‬
‫‪953.45‬‬
‫‪1014.60‬‬
‫‪988.04‬‬
‫‪942.36‬‬
‫‪900.13‬‬
‫‪859.04‬‬
‫‪863.27‬‬
‫‪835.25‬‬
‫‪872.53‬‬

‫‪169‬‬
‫‪170‬‬
‫‪171‬‬
‫‪172‬‬
‫‪173‬‬
‫‪174‬‬
‫‪175‬‬
‫‪176‬‬
‫‪177‬‬
‫‪178‬‬
‫‪179‬‬
‫‪180‬‬

‫‪1‬‬
‫‪2‬‬
‫‪3‬‬
‫‪4‬‬
‫‪5‬‬
‫‪6‬‬
‫‪7‬‬
‫‪8‬‬
‫‪9‬‬
‫‪10‬‬
‫‪11‬‬
‫‪12‬‬

‫ﺗﻤﺮﻳﻦ‪ :‬وﻟﺪ ﺗﻨﺒﺆات ﻹﻧﺘﺎج اﻟﺤﻠﻴﺐ اﻟﻴﻮﻣﻲ ﺑﺈﺳﺘﺨﺪام اﻟﻄﺮﻳﻘﺘﻴﻦ اﻟﻤﻌﻄﺎة وﻗﺎرﻧﻬﺎ ﺑﺎﻟﻘﻴﻢ اﻷﺧﻴﺮة‬
‫اﻟﻨﺎﺗﺠﺔ ﻣﻦ اﻟﺒﺮﻧﺎﻣﺞ‬

‫‪١٦٦‬‬

‫اﻟﻔﺼﻞ اﻟﻌﺎﺷﺮ‬
‫اﻟﺘﻤﻬﻴﺪ و اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪Using Moving‬‬
‫‪Average Smoothing for Forecasting‬‬
‫ﻳﺴﺘﺨﺪم اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻟﺘﻤﻬﻴﺪ اﻟﻤﺸﺎهﺪات وذﻟﻚ ﺑﺘﻘﻠﻴﻞ ﺗﺒﺎﻳﻦ اﻷﺧﻄﺎء ﻓﻤﺜﻼ ﻟﻮ آﺎن ﻟﺪﻳﻨﺎ‬
‫ﻣﺸﺎهﺪات ﻣﻦ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ‪ z1 , z2 , z3 ,K , zn −2 , zn −1 , zn‬ﻓﺎﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ‪m‬‬
‫ﻟﻠﻤﺸﺎهﺪات ﻳﺤﺴﺐ ﻣﻦ اﻟﻌﻼﻗﺔ‬
‫‪1‬‬
‫‪zˆt = ( zt + zt −1 + zt −2 + L + zt −m+1 ) , t = m, m + 1,..., n‬‬
‫‪m‬‬
‫أو‬
‫‪1‬‬
‫‪zˆt = zˆt −1 + ( zt − zt −m ) , t = m, m + 1,..., n‬‬
‫‪m‬‬
‫ﻻﺣﻆ ان ﻋﺪد اﻟﻤﺸﺎهﺪات اﺻﺒﺢ ﺑﻌﺪ اﻟﺘﻤﻬﻴﺪ ‪. n − m + 1‬‬
‫ﻓﻤﺜﻼ ﻟﻮ آﺎﻧﺖ ‪ m=٣‬ﻓﺈن اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻟﺜﺔ هﻮ‬
‫‪1‬‬
‫) ‪( z3 + z2 + z1‬‬
‫‪3‬‬
‫‪1‬‬
‫‪1‬‬
‫) ‪zˆ4 = ( z4 + z3 + z2 ) or zˆ4 = zˆ3 + ( z4 − z1‬‬
‫‪3‬‬
‫‪3‬‬
‫‪M‬‬
‫‪1‬‬
‫‪1‬‬
‫) ‪zˆn = ( zn + zn −1 + zn −2 ) or zˆn = zˆn −1 + ( zn − zn −3‬‬
‫‪3‬‬
‫‪3‬‬
‫= ‪zˆ3‬‬

‫وﻟﻜﻲ ﻧﺮى آﻴﻒ ﻳﻌﻤﻞ اﻟﺘﻤﻬﻴﺪ ﻟﺘﻘﻠﻴﻞ ﺗﺒﺎﻳﻦ اﻷﺧﻄﺎء ﻟﻨﻔﺘﺮض ان اﻟﻤﺸﺎهﺪات ﺗﺘﺒﻊ اﻟﻨﻤﻮذج‬
‫‪WN ( 0, σ 2 ) , t = 1, 2,..., n‬‬

‫ﻓﻴﻜﻮن‬

‫‪zt = µ + at , at‬‬

‫‪V ( zt ) = σ , ∀t‬‬
‫‪2‬‬

‫وﺑﺎﻟﺘﺎﻟﻲ‬
‫‪2‬‬

‫‪σ‬‬

‫= ) ‪V ( zˆt‬‬

‫‪, t = m, m + 1,..., n‬‬
‫‪m‬‬
‫أي ان اﻟﻤﺸﺎهﺪات اﻟﻤﻤﻬﺪة اﺻﺒﺢ ﺗﺒﺎﻳﻨﻬﺎ أﺻﻐﺮ ﺑـ ‪ m‬ﺿﻌﻒ ﻣﻦ اﻟﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ وهﺬا اﻟﺘﻤﻬﻴﺪ‬
‫ﻟﻸﺧﻄﺎء ﻳﻈﻬﺮ أي ﻧﻤﻂ ﻓﻲ اﻟﻤﺘﺴﻠﺴﻠﺔ آﺎن ﻣﺪﻓﻮﻧﺎ او ﻣﻐﻄﻰ ﻣﻦ ﺗﺄﺛﻴﺮ اﻷﺧﻄﺎء‪.‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬ﺗﺆﺧﺬ ‪ m‬داﺋﻤﺎ ﻓﺮدﻳﺔ وذﻟﻚ ﻟﻨﺘﺠﻨﺐ ﺗﻮﺳﻴﻂ اﻟﻘﻴﻢ اﻟﻤﻤﻬﺪة‪.‬‬

‫اﻟﺘﻨﺒﺆ ﺑﺈﺳﺘﺨﺪام اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪:‬‬
‫ﻳﺆﺧﺬ آﻤﺘﻨﺒﺊ ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪:‬‬

‫‪zn ( l ) = zˆn −1 , l > 0‬‬

‫ﻣﺜﺎل‪:‬‬
‫ﺑﺈﺳﺘﺨﺪام اﻟﺤﺰﻣﺔ اﻷﺣﺼﺎﺋﻴﺔ ‪ MINITAB‬ﻧﺤﻤﻞ اﻟﺒﻴﺎﻧﺎت ﻣﻦ ورﻗﺔ اﻟﻌﻤﻞ ‪EMPLOY.MTW‬‬
‫‪'E:\Mtbwin\DATA\EMPLOY.MTW'.‬‬

‫‪MTB > Retrieve‬‬

‫ﻧﻨﻈﺮ ﻣﺎذا ﺗﺤﻮي ﻣﻦ ﻣﺘﻐﻴﺮات‬
‫‪MTB > info‬‬

‫‪١٦٧‬‬

Information on the Worksheet
Column
C1
C2
C3

Count
60
60
60

Name
Trade
Food
Metals

Metals ‫ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺸﺎهﺪات ﻟﻠﻤﺘﻐﻴﺮ‬
Metals
44.2
43.1
42.4
42.9
45.0
49.4
50.5

44.3
42.6
43.1
43.6
45.5
50.0
51.2

44.4
42.4
43.2
44.7
46.2
50.0
50.7

43.4
42.2
42.8
44.5
46.8
49.6
50.3

42.8
41.8
43.0
45.0
47.5
49.9
49.2

44.3
40.1
42.8
44.8
48.3
49.6
48.1

44.4
42.0
42.5
44.9
48.3
50.7

44.8
42.4
42.6
45.2
49.1
50.7

44.4
43.1
42.3
45.2
48.9
50.9

:‫ﻧﺮﺳﻢ هﺬﻩ اﻟﻤﺸﺎهﺪات‬
MTB > TSPlot 'Metals';
SUBC>
Index;
SUBC>
TDisplay 11;
SUBC>
Symbol;
SUBC>
Connect.

Metals

50

45

40
Index

10

20

30

40

50

60

‫ وﻧﻮﺟﺪ ﺗﻨﺒﺆات‬m=٣ ‫ﻧﻄﺒﻖ اﻵن ﺗﻤﻬﻴﺪا ﻟﻬﺬﻩ اﻟﻤﺸﺎهﺪات ﺑﺈﺳﺘﺨﺪام اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ‬
:‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﻪ‬٦ ‫ﻟـ‬
MTB > %MA 'Metals' 3;
SUBC>
Forecasts 6;
SUBC>
Title "Smoothing and Forecasting Metals".
Executing from file: E:\MTBWIN\MACROS\MA.MAC

Moving average
Data
Length
NMissing

Metals
60.0000
0

Moving Average
Length: 3
Accuracy Measures
MAPE: 1.55036

١٦٨

‫‪0.70292‬‬
‫‪0.76433‬‬
‫‪Lower‬‬

‫‪Upper‬‬

‫‪50.9135‬‬
‫‪50.9135‬‬
‫‪50.9135‬‬
‫‪50.9135‬‬
‫‪50.9135‬‬
‫‪50.9135‬‬

‫‪Forecast‬‬

‫‪Period‬‬

‫‪49.2‬‬
‫‪49.2‬‬
‫‪49.2‬‬
‫‪49.2‬‬
‫‪49.2‬‬
‫‪49.2‬‬

‫‪61‬‬
‫‪62‬‬
‫‪63‬‬
‫‪64‬‬
‫‪65‬‬
‫‪66‬‬

‫‪47.4865‬‬
‫‪47.4865‬‬
‫‪47.4865‬‬
‫‪47.4865‬‬
‫‪47.4865‬‬
‫‪47.4865‬‬

‫‪MAD:‬‬
‫‪MSD:‬‬

‫‪Row‬‬
‫‪1‬‬
‫‪2‬‬
‫‪3‬‬
‫‪4‬‬
‫‪5‬‬
‫‪6‬‬

‫‪Smoothing and Forecasting Metals‬‬

‫‪Actual‬‬
‫‪Predicted‬‬

‫‪50‬‬

‫‪Forecast‬‬

‫‪45‬‬
‫‪Moving Average‬‬
‫‪3‬‬

‫‪Metals‬‬

‫‪Actual‬‬
‫‪Predicted‬‬
‫‪Forecast‬‬

‫‪Length:‬‬

‫‪MAPE: 1.55036‬‬
‫‪0.70292‬‬

‫‪MAD:‬‬

‫‪0.76433‬‬

‫‪MSD:‬‬

‫‪40‬‬
‫‪60‬‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻣﻨﺎﻗﺸﺔ اﻟﻨﺘﺎﺋﺞ‬
‫ﻓﻲ اﻟﻤﺜﺎل اﻟﺤﺎﻟﻲ‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪0‬‬

‫‪Time‬‬

‫‪50.3 + 49.2 + 48.1 147.6‬‬
‫=‬
‫‪= 49.2‬‬
‫‪3‬‬
‫‪3‬‬
‫ﺗﺆﺧﺬ اﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟـ ‪ ٦‬اﻟﻤﺴﺘﻘﺒﻠﻴﺔ أي ﻟﻠﻘﻴﻢ ‪ zn +1 , zn +2 ,..., zn+6‬أو ﻓﻲ هﺬا اﻟﻤﺜﺎل‬
‫‪ z61 , z62 ,..., z66‬آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪z60 (1) = z60 ( 2 ) = L = z60 ( 6 ) = 49.2‬‬

‫= ‪zˆ59‬‬

‫ﻟﺤﺴﺎب ﻓﺘﺮات ﺗﻨﺒﺆ ‪ 95%‬ﻧﺤﺴﺐ اﻟﻜﻤﻴﺎت ‪ ⎡⎣ zn ( l ) ± 1.96σˆ ⎤⎦ , l > 0‬أي‬
‫] ˆ‪ [ 49.2 ± 1.96σ‬ﻟﺠﻤﻴﻊ ﻗﻴﻢ اﻟﺘﻨﺒﺆات اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ ،‬ﻧﺄﺧﺬ اﻟﻘﻴﻤﺔ ‪ MSD = 0.76433‬آﻤﻘﺪر ﻟـ‬

‫‪ σ 2‬أي ‪ σˆ 2 = 0.76433‬ﻓﻴﻜﻮن ‪ σˆ = 0.8743‬وﻋﻠﻴﻪ ﺗﻜﻮن ﻓﺘﺮة ﺗﻨﺒﺆ ‪ 95%‬ﻟﺠﻤﻴﻊ اﻟﻘﻴﻢ‬
‫اﻟﻤﺴﺘﻘﺒﻠﻴﺔ هﻲ‪:‬‬
‫]‪⎡⎣ 49.2 ± 1.96 ( 0.8743) ⎤⎦ = [ 49.2 ± 1.7135] = [ 47.4865,50.9135‬‬
‫أي‪:‬‬
‫‪z60+l ∈ [ 47.4865,50.9135] , l > 0 with probability 0.95‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬ﺗﺤﺴﺐ ‪ MSD‬آﺎﻵﺗﻲ‬

‫‪١٦٩‬‬

‫) ‪− zˆi‬‬

‫‪n −1‬‬

‫‪∑( z‬‬

‫‪i‬‬

‫‪i =2‬‬

‫‪n−2‬‬

‫= ˆ‪MSD = σ‬‬
‫‪2‬‬

‫ﺗﻤﺮﻳﻦ‪:‬‬
‫ﻃﺒﻖ ﻣﺘﻮﺳﻄﺎت ﻣﺘﺤﺮآﺔ ﻣﻦ اﻟﺪرﺟﺎت ‪ ٥‬و ‪ ٧‬ﻋﻠﻲ اﻟﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ وﻗﺮر اﻳﻬﺎ اﻓﻀﻞ ﻟﻠﺘﻨﺒﺆ‬
‫ﻋﻦ اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ؟‪.‬‬
‫اﻟﻮﺳﻴﻂ اﻟﺠﺎري ‪Running Median‬‬
‫اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻳﺘﺄﺛﺮ آﺜﻴﺮا ﺑﺎﻟﻤﺸﺎهﺪات اﻟﺨﺎرﺟﺔ ‪ Outliers‬او اﻟﻤﺘﻄﺮﻓﺔ ﻓﺎﻟﻘﻴﻤﺔ اﻟﻤﺘﻄﺮﻓﺔ‬
‫اﻟﻮاﺣﺪة ﺗﺆﺛﺮ ﻋﻠﻲ ‪ m‬ﻣﻦ اﻟﻤﺘﻮﺳﻄﺎت اﻟﻤﺘﺤﺮآﺔ اﻟﻤﺘﺘﺎﻟﻴﺔ ﻓﻤﺜﻼ ﻟﻮآﺎﻧﺖ ﻟﺪﻳﻦ اﻟﻤﺸﺎهﺪات‬
‫‪18‬‬

‫‪13‬‬

‫‪15‬‬

‫‪11‬‬

‫‪12‬‬

‫‪1500‬‬

‫‪10‬‬

‫‪6‬‬

‫‪8‬‬

‫‪9‬‬

‫)‪z(t‬‬
‫‪5‬‬
‫‪7‬‬
‫‪20‬‬

‫‪3‬‬

‫وﻟﻬﺎ اﻟﺸﻜﻞ‬
‫‪1500‬‬

‫‪1000‬‬

‫)‪z(t‬‬
‫‪500‬‬

‫‪0‬‬
‫‪10‬‬

‫‪In d e x‬‬

‫‪5‬‬

‫ﺑﺄﺧﺬ ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ‪ ٣‬ﻧﺠﺪ‬
‫‪M o v in g A v e r a g e‬‬

‫‪A c tu a l‬‬

‫‪1500‬‬

‫‪P re d ic te d‬‬
‫‪A c tu a l‬‬
‫‪P re d ic te d‬‬

‫‪1000‬‬

‫)‪z(t‬‬

‫‪M o v in g A v e ra g e‬‬
‫‪3‬‬
‫‪1081‬‬

‫‪500‬‬

‫‪L e n g th :‬‬
‫‪M APE :‬‬

‫‪273‬‬

‫‪M AD :‬‬

‫‪268811‬‬

‫‪MSD:‬‬

‫‪0‬‬

‫‪15‬‬

‫‪5‬‬

‫‪10‬‬

‫‪0‬‬

‫‪T im e‬‬

‫ﻻﺣﻆ ان اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻨﺎﺗﺞ ﺗﺄﺛﺮت ﻓﻴﺔ ﺛﻼﺛﺔ ﻗﻴﻢ ﺑﺎﻟﻘﻴﻤﺔ اﻟﻤﺘﻄﺮﻓﺔ‪.‬‬
‫ﻟﻠﺘﻐﻠﺐ ﻋﻠﻰ ﻣﺜﻞ هﺬﻩ اﻟﺼﻌﻮﺑﺎت ﻳﺴﺘﺨﺪم اﻟﻮﺳﻴﻂ اﻟﺠﺎري ذا اﻟﻄﻮل اﻟﻔﺮدي آﻤﻤﻬﺪ ﻏﻴﺮ ﺧﻄﻲ‬
‫واﻟﺬي ﻻﻳﺘﺄﺛﺮ ﺑﺎﻗﻴﻢ اﻟﻤﺘﻄﺮﻓﺔ‪.‬‬
‫اﻟﻮﺳﻴﻂ اﻟﺠﺎري ذا اﻟﻄﻮل اﻟﻔﺮدي ‪ j = 2i + 1‬ﻟﻤﺸﺎهﺪات ‪ z1 , z2 , z3 ,K , zn −2 , zn −1 , zn‬ﻳﺤﺴﺐ‬
‫ﻣﻦ اﻟﻌﻼﻗﺔ‬

‫‪z%t = med ( zt −i ,..., zt ,..., zt +i ) , j = 2i + 1‬‬

‫‪١٧٠‬‬

‫ﻓﻤﺜﻼ ﻟﻘﻴﻤﺔ ‪ j = 3‬ﺗﺼﺒﺢ اﻟﻌﻼﻗﺔ ) ‪ z%t = med ( zt −1 , zt , zt +1‬وﺑﺄﺧﺬ وﺳﻴﻂ ﺟﺎري ذا اﻟﻄﻮل ‪٣‬‬
‫ﻟﻠﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ ﻧﺠﺪ‬

‫‪15‬‬

‫)‪smoothz(t‬‬

‫‪10‬‬

‫‪5‬‬

‫‪12‬‬

‫‪8‬‬

‫‪10‬‬

‫‪4‬‬

‫‪6‬‬

‫‪2‬‬

‫‪In d e x‬‬

‫واذا آﺎﻧﺖ اﻟﻘﻴﻤﺔ اﻟﺤﻘﻴﻘﻴﺔ ﻟـ ‪ z9‬هﻲ ‪ ١٥‬وﻟﻴﺲ ‪ ١٥٠٠‬ﻓﺈن اﻟﻤﺸﺎهﺪات اﻟﺤﻘﻴﻘﻴﺔ ﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬

‫‪20‬‬

‫‪15‬‬

‫)‪z(t‬‬
‫‪10‬‬

‫‪5‬‬

‫‪10‬‬

‫ﻗﺎرن ﺑﻴﻦ اﻟﻨﺘﻴﺠﺘﻴﻦ‪.‬‬

‫‪١٧١‬‬

‫‪5‬‬

‫‪Index‬‬

‫اﻟﻔﺼﻞ اﻟﺤﺎدي ﻋﺸﺮ‬
‫اﻟﺘﻤﻬﻴﺪ و اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺒﺴﻴﻂ ‪Using Single‬‬
‫‪: Exponential Smoothing for Forecasting‬‬
‫اﻟﺘﻤﻬﻴﺪ ﺑﻮاﺳﻄﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻳﻌﻄﻲ ﺟﻤﻴﻊ اﻟﺒﻴﺎﻧﺎت ﻧﻔﺲ اﻷهﻤﻴﺔ وﺑﺎﻟﺘﺎﻟﻲ ﻓﺈن اﻟﻘﻴﻢ اﻟﻘﺪﻳﻤﺔ‬
‫ﻧﻮﻋﺎ ﺗﺆﺛﺮ ﻧﻔﺲ اﻟﺘﺄﺛﻴﺮ آﺎﻟﻘﻴﻢ اﻟﺤﺪﻳﺜﺔ وهﺬا ﻗﺪ ﻻﻳﻜﻮن ﻣﻦ اﻟﻨﺎﺣﻴﺔ اﻟﻌﻤﻠﻴﺔ ﺻﺤﻴﺤﺎ‪ ،‬اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ‬
‫ﻋﻠﻰ اﻟﻌﻜﺲ ﻳﻌﻄﻲ اﻟﻘﻴﻢ اﻷآﺜﺮ‬
‫ﺣﺪاﺛﺔ أهﻤﻴﺔ أآﺒﺮ واﻟﻘﻴﻢ اﻻﺧﺮي ﺗﻌﻄﻰ اهﻤﻴﺔ ﺗﺘﻨﺎﻗﺺ اﺳﻴﺎ ﻣﻊ ﻗﺪﻣﻬﺎ‪ .‬ﻓﻤﺜﻼ ﻟﻮ آﺎن ﻟﺪﻳﻨﺎ‬
‫ﻣﺸﺎهﺪات ﻣﻦ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ‪ z1 , z2 , z3 ,K , zn −2 , zn −1 , zn‬ﻓﺎﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ‪m‬‬
‫ﻟﻠﻤﺸﺎهﺪات ﻳﺤﺴﺐ ﻣﻦ اﻟﻌﻼﻗﺔ‬
‫‪1‬‬
‫‪zˆt = ( zt + zt −1 + zt −2 + L + zt −m+1 ) , t = m, m + 1,..., n‬‬
‫‪m‬‬
‫واﻟﺘﻲ ﻳﻤﻜﻦ آﺘﺎﺑﺘﻬﺎ‬
‫‪1‬‬
‫‪1‬‬
‫‪1‬‬
‫‪1‬‬
‫‪zˆt = zt + zt −1 + zt −2 + L + zt −m+1 , t = m, m + 1,..., n‬‬
‫‪m‬‬
‫‪m‬‬
‫‪m‬‬
‫‪m‬‬
‫‪1‬‬
‫= ‪zˆt = β zt + β zt −1 + β zt −2 + L + β zt −m+1 , t = m, m + 1,..., n, β‬‬
‫‪m‬‬
‫أي ان اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻳﻌﻄﻲ ﺟﻤﻴﻊ اﻟﺒﻴﺎﻧﺎت ﻧﻔﺲ اﻟﻮزن ‪β‬‬
‫اﻵن ﻟﻮ أﻋﻄﻴﻨﺎ اﻟﺒﻴﺎﻧﺎت اوزان ﺗﺘﻨﺎﻗﺺ اﺳﻴﺎ ﻣﻊ ﺑُﻌﺪ اﻟﻤﺸﺎهﺪات ﻋﻦ اﻟﻘﻴﻤﺔ اﻟﺤﺎﺿﺮة ‪ zn‬آﺎﻟﺘﺎﻟﻲ‬
‫‪st = α zt + α (1 − α ) zt −1 + α (1 − α ) zt −2 + L , t = 1, 2,..., n, 0 < α < 1‬‬
‫‪2‬‬

‫اﻟﻘﻴﻤﺔ ‪ st‬هﻲ ﻣﺘﻮﺳﻂ ﻣﻮزون ﺑﺄوزان ﺗﺘﻨﺎﻗﺺ اﺳﻴﺎ ﻟﺠﻤﻴﻊ اﻟﻘﻴﻢ اﻟﺴﺎﺑﻘﺔ وهﺬا ﻣﺎﻳﺴﻤﻰ ﺑﺎﻟﺘﻤﻬﻴﺪ‬
‫اﻻﺳﻲ اﻟﺒﺴﻴﻂ وﻳﻜﺘﺐ ﺑﺸﻜﻞ ﺗﻜﺮاري‬

‫‪st = α zt + (1 − α ) st −1 , t = 1, 2,..., n, s0 = z‬‬

‫وﺗﺆﺧﺬ اﻟﺘﻨﺒﺆات‬

‫‪zn ( l ) = sn , l ≥ 1‬‬

‫ﻣﺜﺎل‪:‬‬

‫ﺗﺤﻤﻞ اﻟﺒﻴﺎﻧﺎت ﻣﻦ ورﻗﺔ اﻟﻌﻤﻞ‬

‫‪EMPLOY.MTW‬‬
‫‪'E:\Mtbwin\DATA\EMPLOY.MTW'.‬‬

‫‪MTB > Retrieve‬‬

‫ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺸﺎهﺪات ﻓﻲ اﻟﺘﻐﻴﺮ ‪Metals‬‬
‫‪44.4‬‬
‫‪43.1‬‬
‫‪42.3‬‬
‫‪45.2‬‬
‫‪48.9‬‬
‫‪50.9‬‬

‫‪44.8‬‬
‫‪42.4‬‬
‫‪42.6‬‬
‫‪45.2‬‬
‫‪49.1‬‬
‫‪50.7‬‬

‫‪44.4‬‬
‫‪42.0‬‬
‫‪42.5‬‬
‫‪44.9‬‬
‫‪48.3‬‬
‫‪50.7‬‬

‫‪44.3‬‬
‫‪40.1‬‬
‫‪42.8‬‬
‫‪44.8‬‬
‫‪48.3‬‬
‫‪49.6‬‬
‫‪48.1‬‬

‫‪42.8‬‬
‫‪41.8‬‬
‫‪43.0‬‬
‫‪45.0‬‬
‫‪47.5‬‬
‫‪49.9‬‬
‫‪49.2‬‬

‫‪43.4‬‬
‫‪42.2‬‬
‫‪42.8‬‬
‫‪44.5‬‬
‫‪46.8‬‬
‫‪49.6‬‬
‫‪50.3‬‬

‫‪44.4‬‬
‫‪42.4‬‬
‫‪43.2‬‬
‫‪44.7‬‬
‫‪46.2‬‬
‫‪50.0‬‬
‫‪50.7‬‬

‫‪44.3‬‬
‫‪42.6‬‬
‫‪43.1‬‬
‫‪43.6‬‬
‫‪45.5‬‬
‫‪50.0‬‬
‫‪51.2‬‬

‫‪Metals‬‬
‫‪44.2‬‬
‫‪43.1‬‬
‫‪42.4‬‬
‫‪42.9‬‬
‫‪45.0‬‬
‫‪49.4‬‬
‫‪50.5‬‬

‫ﻧﺮﺳﻢ هﺬﻩ اﻟﻤﺸﺎهﺪات‪:‬‬
‫;'‪MTB > TSPlot 'Metals‬‬
‫>‪SUBC‬‬
‫;‪Index‬‬
‫>‪SUBC‬‬
‫;‪TDisplay 11‬‬
‫>‪SUBC‬‬
‫;‪Symbol‬‬

‫‪١٧٢‬‬

SUBC>

Connect.

M
etals

50

45

40
Index

10

20

30

40

50

60

α = 0.2 ‫ﻧﻄﺒﻖ اﻵن ﺗﻤﻬﻴﺪا ﻟﻬﺬﻩ اﻟﻤﺸﺎهﺪات ﺑﺈﺳﺘﺨﺪام اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺒﺴﻴﻂ وﻧﺄﺧﺬ ﺛﺎﺑﺖ ﺗﻤﻬﻴﺪ‬
:‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﻪ‬٦ ‫وﻧﻮﺟﺪ ﺗﻨﺒﺆات ﻟـ‬
MTB > %SES 'Metals';
SUBC>
Weight 0.2;
SUBC>
Forecasts 6;
SUBC>
Title "Smoothing and Forecasting Metals";
SUBC>
Initial 6.
Single Exponential Smoothing
Data
Metals
Length
60.0000
NMissing
0
Smoothing Constant
Alpha: 0.2
Accuracy Measures
MAPE: 2.17304
MAD: 1.00189
MSD: 1.45392
Row

Period

Forecast

Lower

Upper

1
2
3
4
5
6

61
62
63
64
65
66

49.7216
49.7216
49.7216
49.7216
49.7216
49.7216

47.2670
47.2670
47.2670
47.2670
47.2670
47.2670

52.1763
52.1763
52.1763
52.1763
52.1763
52.1763

Smoothing and Forecasting Metals

Actual
Predicted
Forecast

Metals

50

Actual
Predicted
Forecast

45

Smoothing Constant
Alpha:

40
0

10

20

30

Time

١٧٣

40

50

60

0.200

MAPE:

2.17304

MAD:

1.00189

MSD:

1.45392

‫ﺛﺎﻧﻴﺎ‪ :‬ﻣﻨﺎﻗﺸﺔ اﻟﻨﺘﺎﺋﺞ‬
‫‪ -١‬ﻳﺤﺴﺐ اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺒﺴﻴﻂ ﻟﻤﺸﺎهﺪات‬
‫ﻣﻦ اﻟﻌﻼﻗﺔ اﻟﺘﻜﺮارﻳﺔ‪:‬‬

‫‪ z1 , z2 ,K , zn −1 , zn −2‬ﺑﺜﺎﺑﺖ ﺗﻤﻬﻴﺪ ‪α = 0.2‬‬

‫‪si = α zi + (1 − α ) si −1 , i = 1, 2,..., n‬‬
‫ﻟﻜﻲ ﻧﺒﺪأ اﻟﻌﻼﻗﺔ اﻟﺘﻜﺮارﻳﺔ ﻟﺤﺴﺎب اﻟﻘﻴﻢ اﻟﻤﻤﻬﺪة اﺳﻴﺎ ﻧﺤﺘﺎج اﻟﻲ اﻟﻘﻴﻤﺔ اﻻوﻟﻴﺔ ‪ s0‬واﻟﺘﻲ ﺗﺤﺴﺐ‬
‫ﺑﻌﺪة ﻃﺮق‪ ،‬أﺣﺪ هﺬﻩ اﻟﻄﺮق واﻟﺘﻲ ﺳﻨﺴﺘﺨﺪﻣﻬﺎ هﻲ وﺿﻊ ‪ s0‬ﻣﺴﺎوﻳﺔ ﻟﻠﻤﺘﻮﺳﻂ‬
‫‪m‬‬

‫) ‪, m = 6 ( or n, if n<6‬‬

‫‪∑z‬‬

‫‪i‬‬

‫‪i =1‬‬

‫‪m‬‬

‫= ‪ s0‬ﻓﻔﻲ ﻣﺜﺎﻟﻨﺎ‬

‫‪44.2 + 44.3 + 44.4 + 43.4 + 42.8 + 44.3‬‬
‫‪= 43.9‬‬
‫‪6‬‬

‫= ‪s0‬‬

‫وﺑﺎﻟﺘﺎﻟﻲ ﻳﻜﻮن‬
‫‪s1 = α z1 + (1 − α ) s0 = 0.2 ( 44.2 ) + 0.8 ( 43.9 ) = 8.84 + 35.12 = 43.96‬‬
‫‪s2 = α z2 + (1 − α ) s1 = 0.2 ( 44.3) + 0.8 ( 43.96 ) = 8.86 + 35.168 = 44.028‬‬
‫وهﻜﺬا ﻧﺴﺘﻤﺮ ﺣﺘﻲ ﺁﺧﺮ ﻣﺸﺎهﺪة ﻓﻴﻨﺘﺞ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪RESI1‬‬

‫‪FITS1‬‬

‫‪SMOO1‬‬

‫‪Metals‬‬

‫‪Time‬‬

‫‪0.30000‬‬
‫‪0.34000‬‬
‫‪0.37200‬‬
‫‪-0.70240‬‬
‫‪-1.16192‬‬
‫‪0.57046‬‬
‫‪0.55637‬‬
‫‪0.84510‬‬
‫‪0.27608‬‬
‫‪-1.07914‬‬
‫‪-1.36331‬‬
‫‪-1.29065‬‬
‫‪-1.23252‬‬
‫‪-1.38601‬‬
‫‪-2.80881‬‬
‫‪-0.34705‬‬
‫‪0.12236‬‬
‫‪0.79789‬‬
‫‪-0.06169‬‬
‫‪0.65065‬‬
‫‪0.62052‬‬
‫‪0.09642‬‬
‫‪0.27713‬‬
‫‪0.02171‬‬
‫‪-0.28264‬‬
‫‪-0.12611‬‬
‫‪-0.40089‬‬
‫‪0.27929‬‬
‫‪0.92343‬‬
‫‪1.83875‬‬
‫‪1.27100‬‬
‫‪1.51680‬‬
‫‪1.01344‬‬
‫‪0.91075‬‬
‫‪1.02860‬‬

‫‪43.9000‬‬
‫‪43.9600‬‬
‫‪44.0280‬‬
‫‪44.1024‬‬
‫‪43.9619‬‬
‫‪43.7295‬‬
‫‪43.8436‬‬
‫‪43.9549‬‬
‫‪44.1239‬‬
‫‪44.1791‬‬
‫‪43.9633‬‬
‫‪43.6906‬‬
‫‪43.4325‬‬
‫‪43.1860‬‬
‫‪42.9088‬‬
‫‪42.3470‬‬
‫‪42.2776‬‬
‫‪42.3021‬‬
‫‪42.4617‬‬
‫‪42.4494‬‬
‫‪42.5795‬‬
‫‪42.7036‬‬
‫‪42.7229‬‬
‫‪42.7783‬‬
‫‪42.7826‬‬
‫‪42.7261‬‬
‫‪42.7009‬‬
‫‪42.6207‬‬
‫‪42.6766‬‬
‫‪42.8613‬‬
‫‪43.2290‬‬
‫‪43.4832‬‬
‫‪43.7866‬‬
‫‪43.9892‬‬
‫‪44.1714‬‬

‫‪43.9600‬‬
‫‪44.0280‬‬
‫‪44.1024‬‬
‫‪43.9619‬‬
‫‪43.7295‬‬
‫‪43.8436‬‬
‫‪43.9549‬‬
‫‪44.1239‬‬
‫‪44.1791‬‬
‫‪43.9633‬‬
‫‪43.6906‬‬
‫‪43.4325‬‬
‫‪43.1860‬‬
‫‪42.9088‬‬
‫‪42.3470‬‬
‫‪42.2776‬‬
‫‪42.3021‬‬
‫‪42.4617‬‬
‫‪42.4494‬‬
‫‪42.5795‬‬
‫‪42.7036‬‬
‫‪42.7229‬‬
‫‪42.7783‬‬
‫‪42.7826‬‬
‫‪42.7261‬‬
‫‪42.7009‬‬
‫‪42.6207‬‬
‫‪42.6766‬‬
‫‪42.8613‬‬
‫‪43.2290‬‬
‫‪43.4832‬‬
‫‪43.7866‬‬
‫‪43.9892‬‬
‫‪44.1714‬‬
‫‪44.3771‬‬

‫‪44.2‬‬
‫‪44.3‬‬
‫‪44.4‬‬
‫‪43.4‬‬
‫‪42.8‬‬
‫‪44.3‬‬
‫‪44.4‬‬
‫‪44.8‬‬
‫‪44.4‬‬
‫‪43.1‬‬
‫‪42.6‬‬
‫‪42.4‬‬
‫‪42.2‬‬
‫‪41.8‬‬
‫‪40.1‬‬
‫‪42.0‬‬
‫‪42.4‬‬
‫‪43.1‬‬
‫‪42.4‬‬
‫‪43.1‬‬
‫‪43.2‬‬
‫‪42.8‬‬
‫‪43.0‬‬
‫‪42.8‬‬
‫‪42.5‬‬
‫‪42.6‬‬
‫‪42.3‬‬
‫‪42.9‬‬
‫‪43.6‬‬
‫‪44.7‬‬
‫‪44.5‬‬
‫‪45.0‬‬
‫‪44.8‬‬
‫‪44.9‬‬
‫‪45.2‬‬

‫‪1‬‬
‫‪2‬‬
‫‪3‬‬
‫‪4‬‬
‫‪5‬‬
‫‪6‬‬
‫‪7‬‬
‫‪8‬‬
‫‪9‬‬
‫‪10‬‬
‫‪11‬‬
‫‪12‬‬
‫‪13‬‬
‫‪14‬‬
‫‪15‬‬
‫‪16‬‬
‫‪17‬‬
‫‪18‬‬
‫‪19‬‬
‫‪20‬‬
‫‪21‬‬
‫‪22‬‬
‫‪23‬‬
‫‪24‬‬
‫‪25‬‬
‫‪26‬‬
‫‪27‬‬
‫‪28‬‬
‫‪29‬‬
‫‪30‬‬
‫‪31‬‬
‫‪32‬‬
‫‪33‬‬
‫‪34‬‬
‫‪35‬‬

‫‪١٧٤‬‬

‫‪0.82288‬‬
‫‪0.45830‬‬
‫‪0.86664‬‬
‫‪1.39331‬‬
‫‪1.71465‬‬
‫‪2.07172‬‬
‫‪2.45738‬‬
‫‪1.96590‬‬
‫‪2.37272‬‬
‫‪1.69818‬‬
‫‪1.85854‬‬
‫‪2.08683‬‬
‫‪1.66947‬‬
‫‪0.93557‬‬
‫‪1.04846‬‬
‫‪0.53877‬‬
‫‪1.53101‬‬
‫‪1.22481‬‬
‫‪1.17985‬‬
‫‪0.54388‬‬
‫‪1.13510‬‬
‫‪0.40808‬‬
‫‪-0.07353‬‬
‫‪-1.15883‬‬
‫‪-2.02706‬‬

‫‪44.3771‬‬
‫‪44.5417‬‬
‫‪44.6334‬‬
‫‪44.8067‬‬
‫‪45.0853‬‬
‫‪45.4283‬‬
‫‪45.8426‬‬
‫‪46.3341‬‬
‫‪46.7273‬‬
‫‪47.2018‬‬
‫‪47.5415‬‬
‫‪47.9132‬‬
‫‪48.3305‬‬
‫‪48.6644‬‬
‫‪48.8515‬‬
‫‪49.0612‬‬
‫‪49.1690‬‬
‫‪49.4752‬‬
‫‪49.7202‬‬
‫‪49.9561‬‬
‫‪50.0649‬‬
‫‪50.2919‬‬
‫‪50.3735‬‬
‫‪50.3588‬‬
‫‪50.1271‬‬

‫‪44.5417‬‬
‫‪44.6334‬‬
‫‪44.8067‬‬
‫‪45.0853‬‬
‫‪45.4283‬‬
‫‪45.8426‬‬
‫‪46.3341‬‬
‫‪46.7273‬‬
‫‪47.2018‬‬
‫‪47.5415‬‬
‫‪47.9132‬‬
‫‪48.3305‬‬
‫‪48.6644‬‬
‫‪48.8515‬‬
‫‪49.0612‬‬
‫‪49.1690‬‬
‫‪49.4752‬‬
‫‪49.7202‬‬
‫‪49.9561‬‬
‫‪50.0649‬‬
‫‪50.2919‬‬
‫‪50.3735‬‬
‫‪50.3588‬‬
‫‪50.1271‬‬
‫‪49.7216‬‬

‫‪45.2‬‬
‫‪45.0‬‬
‫‪45.5‬‬
‫‪46.2‬‬
‫‪46.8‬‬
‫‪47.5‬‬
‫‪48.3‬‬
‫‪48.3‬‬
‫‪49.1‬‬
‫‪48.9‬‬
‫‪49.4‬‬
‫‪50.0‬‬
‫‪50.0‬‬
‫‪49.6‬‬
‫‪49.9‬‬
‫‪49.6‬‬
‫‪50.7‬‬
‫‪50.7‬‬
‫‪50.9‬‬
‫‪50.5‬‬
‫‪51.2‬‬
‫‪50.7‬‬
‫‪50.3‬‬
‫‪49.2‬‬
‫‪48.1‬‬

‫‪36‬‬
‫‪37‬‬
‫‪38‬‬
‫‪39‬‬
‫‪40‬‬
‫‪41‬‬
‫‪42‬‬
‫‪43‬‬
‫‪44‬‬
‫‪45‬‬
‫‪46‬‬
‫‪47‬‬
‫‪48‬‬
‫‪49‬‬
‫‪50‬‬
‫‪51‬‬
‫‪52‬‬
‫‪53‬‬
‫‪54‬‬
‫‪55‬‬
‫‪56‬‬
‫‪57‬‬
‫‪58‬‬
‫‪59‬‬
‫‪60‬‬

‫اﻟﻌﻤﻮد اﻟﺮاﺑﻊ ‪ SMOO١‬ﻳﺤﻮى اﻟﻘﻴﻢ اﻟﻤﻤﻬﺪة أي ‪ si , i = 1, 2,...,60‬اﻟﻌﻤﻮد اﻟﺨﺎﻣﺲ ‪FITS١‬‬
‫ﻳﺤﻮى اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ أي ‪ zˆi = si −1 , i = 1, 2,...,60‬اﻟﻌﻤﻮد اﻟﺨﺎﻣﺲ ‪ RESI١‬ﻳﺤﻮي اﻷﺧﻄﺎء‬
‫)اﻟﺒﻮاﻗﻲ ‪ (Residuals‬أي ‪ei = zi − zˆi , i = 1, 2,...,60‬‬
‫‪ -٢‬ﻳﺆﺧﺬ آﻤﺘﻨﺒﺊ ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺁﺧﺮ ﻗﻴﻤﺔ ﻣﻤﻬﺪة أي‪:‬‬
‫‪z n ( l ) = sn , l > 0‬‬
‫ﻓﻔﻲ اﻟﻤﺜﺎل اﻟﺤﺎﻟﻲ‬
‫‪z60 ( l ) = 49.7216, l > 0‬‬
‫ﺗﺆﺧﺬ اﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟـ ‪ ٦‬اﻟﻤﺴﺘﻘﺒﻠﻴﺔ أي ﻟﻠﻘﻴﻢ ‪ zn +1 , zn +2 ,..., zn +6‬أو ﻓﻲ هﺬا اﻟﻤﺜﺎل‬
‫‪ z61 , z62 ,..., z66‬آﺎﻟﺘﺎﻟﻲ‪:‬‬
‫‪z60 (1) = z60 ( 2 ) = L = z60 ( 6 ) = 49.7216‬‬

‫‪ -٣‬ﻟﺤﺴﺎب ﻓﺘﺮات ﺗﻨﺒﺆ ‪ 95%‬ﻧﺤﺴﺐ اﻟﻜﻤﻴﺎت ‪ ⎡⎣ zn ( l ) ± 1.96σˆ ⎤⎦ , l > 0‬أي‬
‫] ˆ‪ [ 49.7216 ± 1.96σ‬ﻟﺠﻤﻴﻊ ﻗﻴﻢ اﻟﺘﻨﺒﺆات اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ ،‬ﻧﺄﺧﺬ اﻟﻘﻴﻤﺔ ‪ MSD = 1.45392‬آﻤﻘﺪر‬

‫ﻟـ ‪ σ 2‬أي ‪ σˆ 2 = 1.45392‬ﻓﻴﻜﻮن ‪ σˆ = 1.205786‬وﻋﻠﻴﻪ ﺗﻜﻮن ﻓﺘﺮة ﺗﻨﺒﺆ ‪ 95%‬ﻟﺠﻤﻴﻊ‬
‫اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ هﻲ‪:‬‬
‫]‪⎡⎣ 49.7216 ± 1.96 (1.205786 ) ⎤⎦ = [ 49.7216 ± 2.3633] = [ 47.35826,52.08494‬‬
‫أي‪:‬‬
‫‪z60+l ∈ [ 47.3582,50.0849] , l > 0 with probability 0.95‬‬
‫ﻣﻼﺣﻈﺔ‪ :‬ﺗﺤﺴﺐ ‪ MSD‬آﺎﻵﺗﻲ‬

‫‪١٧٥‬‬

‫) ‪− zˆi‬‬

‫‪n‬‬

‫‪∑( z‬‬

‫‪i‬‬

‫‪n −1‬‬

‫‪i =1‬‬

‫= ˆ‪MSD = σ‬‬
‫‪2‬‬

‫ﺗﻤﺮﻳﻦ‪:‬‬
‫ﻃﺒﻖ ﺗﻤﻬﻴﺪ اﺳﻲ ﺑﺴﻴﻂ ﻋﻠﻲ اﻟﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ ﻣﺴﺘﺨﺪﻣﺎ ‪ α = 0.3,0.4,0.5‬وﻗﺮر اﻳﻬﺎ اﻓﻀﻞ‬
‫ﻟﻠﺘﻨﺒﺆ ﻋﻦ اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ؟‪.‬‬

‫‪١٧٦‬‬

‫اﻟﻔﺼﻞ اﻟﺜﺎﻧﻲ ﻋﺸﺮ‬
‫اﻟﺘﻤﻬﻴﺪ و اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﻤﺰدوج ‪Using Double‬‬
‫‪: Exponential Smoothing for Forecasting‬‬
‫أوﻻ‪ :‬ﻃﺮﻳﻘﺔ ﺑﺮاون ‪:Brown’s Method‬‬

‫ﻟﻤﺸﺎهﺪات ‪ z1 , z2 ,K , zn −1 , zn −2‬وﻟﺜﺎﺑﺖ ﺗﻤﻬﻴﺪ ‪ 0 < α < 1‬ﻧﻮﺟﺪ اﻟﺘﺎﻟﻲ‪:‬‬

‫‪s = α zt + (1 − α ) st(−1) , t = 1,2,..., n‬‬
‫)‪(1‬‬
‫‪t‬‬

‫‪1‬‬

‫ﺣﻴﺚ ) (‪ st‬ﺗﻤﻬﻴﺪ اﺳﻲ ﺑﺴﻴﻂ ‪) st( ) = st‬اﻧﻈﺮ اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺒﺴﻴﻂ( و )‪ (1‬ﺗﺮﻣﺰ اﻟﻲ درﺟﺔ هﺬا‬
‫اﻟﺘﻤﻬﻴﺪ‬
‫)‪(2‬‬
‫)‪(1‬‬
‫)‪(2‬‬
‫‪st = α st + (1 − α ) st −1 , t = 1,2,..., n‬‬
‫‪1‬‬

‫‪1‬‬

‫ﺣﻴﺚ ) (‪ st‬ﺗﻤﻬﻴﺪ اﺳﻲ ﻣﺰدوج و ) ‪ ( 2‬ﺗﺮﻣﺰ اﻟﻲ درﺟﺔ هﺬا اﻟﺘﻤﻬﻴﺪ‬
‫‪2‬‬

‫‪at = 2 st( ) − st( ) , t = 1, 2,..., n‬‬
‫‪1‬‬

‫‪2‬‬

‫‪α‬‬
‫‪1‬‬
‫‪2‬‬
‫‪st( ) − st( ) , t = 1, 2,..., n‬‬
‫‪1−α‬‬

‫)‬

‫ﺗﺤﺴﺐ اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﻣﻦ اﻟﻤﻌﺎدﻟﺔ‬

‫‪t = 1, 2,..., n‬‬

‫وﺗﺤﺴﺐ اﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +l , l > 0‬ﻣﻦ‬

‫‪zˆt = at + bt t ,‬‬

‫‪zn ( l ) = an + bn l, l > 0‬‬

‫ﺣﺴﺎب اﻟﻘﻴﻢ اﻻوﻟﻴﺔ ) (‪ s0‬و ) (‪ : s0‬ﻣﻦ اﻟﻌﻼﻗﺎت اﻟﺴﺎﺑﻘﺔ ﻧﺠﺪ‬
‫‪1‬‬

‫(‬

‫= ‪bt‬‬

‫‪2‬‬

‫‪b0‬‬
‫‪b0‬‬
‫ﻧﻮﺟﺪ ‪ a0‬و ‪ b0‬ﺑﺈﻧﺤﺪار اﻟﻤﺸﺎهﺪات ﻋﻠﻲ اﻟﺰﻣﻦ ‪t = 1, 2,..., n‬‬
‫ˆ‪ a0 = α‬و ˆ‪b0 = β‬‬

‫‪1−α‬‬

‫‪α‬‬

‫‪1−α‬‬

‫‪α‬‬

‫‪s0( ) = a0 −‬‬
‫‪1‬‬

‫‪s0( ) = a0 − 2‬‬
‫‪2‬‬

‫‪ zt = α + β t + et ,‬وﻳﻜﻮن‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻃﺮﻳﻘﺔ هﻮﻟﺖ ‪:Holt’s Method‬‬

‫ﻟﻤﺸﺎهﺪات ‪ z1 , z2 ,K , zn −1 , zn −2‬وﻟﺜﺎﺑﺘﻲ ﺗﻤﻬﻴﺪ ‪ 0 < α < 1‬و ‪ 0 < γ < 1‬ﻧﻮﺟﺪ اﻟﺘﺎﻟﻲ‪:‬‬
‫‪st = α zt + (1 − α )( st −1 + bt −1 ) , t = 1, 2,..., n‬‬
‫‪t = 1, 2,..., n‬‬
‫ﻧﺤﺴﺐ اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﻣﻦ‬

‫‪bt = γ ( st − st −1 ) + (1 − γ ) bt −1 ,‬‬
‫‪t = 1, 2,..., n‬‬

‫واﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻣﻦ‬

‫‪zˆt = st + bt t ,‬‬

‫‪zn ( l ) = sn + bn l, l > 0‬‬

‫ﻧﺤﺴﺐ اﻟﻘﻴﻢ اﻻوﻟﻴﺔ ‪ s0‬و ‪ b0‬ﻣﻦ‬
‫‪١٧٧‬‬

s0 = z1
b0 = z2 − z1

b0 =

or

( z2 − z1 ) + ( z3 − z2 ) = ( z3 − z1 )

or
2
2
( z − z ) + ( z3 − z2 ) + ( z4 − z3 ) ( z4 − z1 )
b0 = 2 1
=
3
3
:‫ﻣﺜﺎل‬
EMPLOY.MTW ‫ﺗﺤﻤﻞ اﻟﺒﻴﺎﻧﺎت ﻣﻦ ورﻗﺔ اﻟﻌﻤﻞ‬
MTB > Retrieve 'E:\Mtbwin\DATA\EMPLOY.MTW'.

Metals ‫ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺸﺎهﺪات ﻓﻲ اﻟﺘﻐﻴﺮ‬
Metals
44.2
43.1
42.4
42.9
45.0
49.4
50.5

44.3
42.6
43.1
43.6
45.5
50.0
51.2

44.4
42.4
43.2
44.7
46.2
50.0
50.7

43.4
42.2
42.8
44.5
46.8
49.6
50.3

42.8
41.8
43.0
45.0
47.5
49.9
49.2

44.3
40.1
42.8
44.8
48.3
49.6
48.1

44.4
42.0
42.5
44.9
48.3
50.7

44.8
42.4
42.6
45.2
49.1
50.7

44.4
43.1
42.3
45.2
48.9
50.9

:‫ﻧﺮﺳﻢ هﺬﻩ اﻟﻤﺸﺎهﺪات‬
MTB > TSPlot 'Metals';
SUBC>
Index;
SUBC>
TDisplay 11;
SUBC>
Symbol;
SUBC>
Connect.

Metals

50

45

40
Index

10

20

30

١٧٨

40

50

60

‫ ﻧﺴﺘﺨﺪم اﻵن‬،‫ﻧﻄﺒﻖ اﻵن ﺗﻤﻬﻴﺪا ﻟﻬﺬﻩ اﻟﻤﺸﺎهﺪات ﺑﺈﺳﺘﺨﺪام اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﻤﺰدوج ﺑﻄﺮﻳﻘﺔ ﺑﺮاون‬
‫ ﺑﺄوزان ﻣﺘﺴﺎوﻳﺔ ﺳﻮف ﻧﺄﺧﺬهﺎ‬WEIGHT ‫ ﻣﻊ اﻻﻣﺮ اﻟﻔﺮﻋﻲ‬%DES (Macro) ‫اﻟﺒﺮﻣﺞ‬
0.2
MTB > %DES 'Metals';
SUBC>
Weight 0.2 0.2;
SUBC>
Forecasts 6;
SUBC>
Title "Brown's Double Exponential Smoothing";
SUBC>
Table.

Double Exponential Smoothing
Data
Length
NMissing

Metals
60.0000
0

Smoothing Constants
Alpha (level): 0.2
Gamma (trend): 0.2
Accuracy Measures
MAPE: 2.16187
MAD: 0.97032
MSD: 1.62936

Time Metals
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

44.2
44.3
44.4
43.4
42.8
44.3
44.4
44.8
44.4
43.1
42.6
42.4
42.2
41.8
40.1
42.0
42.4
43.1
42.4
43.1
43.2
42.8
43.0
42.8
42.5
42.6
42.3
42.9
43.6
44.7

Smooth

Predict

Error

41.7739
42.4976
43.1686
43.5546
43.7373
44.1459
44.4990
44.8575
45.0620
44.9391
44.6673
44.3271
43.9378
43.4769
42.7011
42.3565
42.1465
42.1286
42.0132
42.0763
42.1877
42.2374
42.3396
42.4078
42.4181
42.4495
42.4207
42.5129
42.7420
43.1797

41.1674
42.0470
42.8607
43.5933
43.9716
44.1074
44.5238
44.8719
45.2275
45.3989
45.1841
44.8088
44.3723
43.8962
43.3514
42.4456
42.0831
41.8857
41.9164
41.8204
41.9347
42.0967
42.1745
42.3098
42.3976
42.4119
42.4509
42.4161
42.5276
42.7996

3.03257
2.25303
1.53927
-0.19330
-1.17163
0.19257
-0.12377
-0.07189
-0.82751
-2.29891
-2.58407
-2.40884
-2.17229
-2.09617
-3.25142
-0.44557
0.31694
1.21426
0.48355
1.27964
1.26533
0.70327
0.82549
0.49024
0.10244
0.18809
-0.15090
0.48394
1.07245
1.90036

١٧٩

31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

44.5
45.0
44.8
44.9
45.2
45.2
45.0
45.5
46.2
46.8
47.5
48.3
48.3
49.1
48.9
49.4
50.0
50.0
49.6
49.9
49.6
50.7
50.7
50.9
50.5
51.2
50.7
50.3
49.2
48.1

43.5507
43.9854
44.3338
44.6511
44.9749
45.2430
45.4157
45.6373
45.9491
46.3285
46.7909
47.3492
47.8339
48.4002
48.8413
49.2966
49.7849
50.1841
50.4162
50.6292
50.7104
50.9509
51.1334
51.3019
51.3407
51.4782
51.4770
51.3649
51.0127
50.4384

43.3133
43.7317
44.2172
44.5889
44.9187
45.2538
45.5197
45.6716
45.8863
46.2106
46.6136
47.1115
47.7173
48.2253
48.8267
49.2707
49.7311
50.2302
50.6202
50.8114
50.9880
51.0137
51.2417
51.4024
51.5509
51.5477
51.6712
51.6311
51.4659
51.0230

1.18668
1.26826
0.58280
0.31112
0.28133
-0.05376
-0.51967
-0.17162
0.31368
0.58938
0.88637
1.18850
0.58266
0.87469
0.07332
0.12930
0.26890
-0.23017
-1.02022
-0.91145
-1.38797
-0.31368
-0.54169
-0.50244
-1.05093
-0.34770
-0.97120
-1.33115
-2.26587
-2.92300

Row

Period

Forecast

Lower

Upper

1
2
3
4
5
6

61
62
63
64
65
66

50.3318
50.2252
50.1186
50.0120
49.9054
49.7988

47.9545
47.7984
47.6384
47.4749
47.3080
47.1381

52.7091
52.6520
52.5987
52.5490
52.5027
52.4594

Brown's Double Exponential Smoothing

Actual
Predicted
Forecast

Metals

50

Actual
Predicted
Forecast

45

Smoothing Constants
Alpha (level): 0.200
Gamma (trend):0.200
MAPE:
MAD:
MSD:

40
0

10

20

30

Time

١٨٠

40

50

60

2.16187
0.97032
1.62936

: b0 ‫ و‬a0 ‫إﻳﺠﺎد‬
MTB >
DATA>
DATA>
MTB >

set c4
1:60
end
regr c3 1 c4

Regression Analysis
The regression equation is
Metals = 41.0 + 0.152 C4

1−α

‫ وﻣﻨﻬﺎ ﻧﺤﺴﺐ‬b0 = 0.152 ‫ و‬a0 = 41.0 ‫إذا‬

0.8
( 0.152 ) = 41.608
α
0.2
1−α
⎛ 0.8 ⎞
2
s0( ) = a0 − 2
b0 = 41.0 − 2 ⎜
⎟ ( 0.152 ) = 42.216
α
⎝ 0.2 ⎠
1
s1( ) = ( 0.2 )( 44.2 ) + ( 0.8 )( 41.608 ) = 42.1264
s0( ) = a0 −
1

b0 = 41.0 −

s1( ) = ( 0.2 )( 42.1264 ) + ( 0.8 )( 42.216 ) = 42.19808
2

a1 = ( 2 )( 42.1264 ) − 42.19808 = 42.05472

( 0.2 )
( 42.19808 − 42.05472 ) = 0.03584
( 0.8)
zˆ1 = 42.05472 + ( 0.03584 )(1) = 42.09056

b1 =

… ‫وهﻜﺬا اﻟﺦ‬
.‫ آﻤﺎ ﻓﻲ اﻷﻣﺜﻠﺔ اﻟﺴﺎﺑﻘﺔ‬MSD ‫ﺗﺤﺴﺐ ﻓﺘﺮات اﻟﺘﻨﺒﺆ ﺑﺈﺳﺘﺨﺪام‬
:‫ﻣﺜﺎل‬
WEIGHT ‫ ﻣﻊ اﻻﻣﺮ اﻟﻔﺮﻋﻲ‬%DES (Macro) ‫ﻟﺘﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ هﻮﻟﺖ ﻧﺴﺘﺨﺪم اﻵن اﻟﺒﺮﻣﺞ‬
γ = 0.3 ‫ و‬α = 0.2 ‫ﺑﺄوزان ﻣﺨﺘﻠﻔﺔ ﺳﻮف ﻧﺄﺧﺬ‬
MTB > RETR 'E:\Mtbwin\DATA\EMPLOY.MTW'.
Retrieving worksheet from file: E:\Mtbwin\DATA\EMPLOY.MTW
Worksheet was saved on 6/ 5/1996
MTB > %DES 'Metals';
SUBC>
Weight 0.2 0.3;
SUBC>
Forecasts 6;
SUBC>
Title "Holt's Double Exponential Smoothing";
SUBC>
Table.

Double Exponential Smoothing
Data
Length
NMissing

Metals
60.0000
0

١٨١

Smoothing Constants
Alpha (level): 0.2
Gamma (trend): 0.3
Accuracy Measures
MAPE: 2.15656
MAD: 0.96328
MSD: 1.56274

Time

Metals

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48

44.2
44.3
44.4
43.4
42.8
44.3
44.4
44.8
44.4
43.1
42.6
42.4
42.2
41.8
40.1
42.0
42.4
43.1
42.4
43.1
43.2
42.8
43.0
42.8
42.5
42.6
42.3
42.9
43.6
44.7
44.5
45.0
44.8
44.9
45.2
45.2
45.0
45.5
46.2
46.8
47.5
48.3
48.3
49.1
48.9
49.4
50.0
50.0

Smooth
41.7739
42.5461
43.2891
43.7501
43.9779
44.3895
44.7334
45.0685
45.2405
45.0676
44.7113
44.2595
43.7466
43.1634
42.2751
41.8139
41.5361
41.5057
41.4371
41.5799
41.8054
41.9895
42.2254
42.4205
42.5395
42.6522
42.6793
42.7982
43.0394
43.4861
43.8762
44.3257
44.6858
45.0007
45.3066
45.5449
45.6749
45.8384
46.0888
46.4159
46.8406
47.3799
47.8665
48.4419
48.9016
49.3693
49.8653
50.2702

Predict
41.1674
42.1076
43.0113
43.8376
44.2724
44.4118
44.8167
45.1356
45.4506
45.5595
45.2391
44.7244
44.1332
43.5043
42.8188
41.7674
41.3202
41.1072
41.1964
41.1999
41.4568
41.7869
42.0317
42.3257
42.5493
42.6653
42.7741
42.7728
42.8993
43.1826
43.7202
44.1571
44.6573
45.0259
45.3333
45.6312
45.8436
45.9230
46.0610
46.3199
46.6757
47.1499
47.7582
48.2774
48.9020
49.3617
49.8317
50.3378

Error
3.03257
2.19238
1.38868
-0.43760
-1.47237
-0.11184
-0.41671
-0.33560
-1.05057
-2.45952
-2.63911
-2.32443
-1.93322
-1.70426
-2.71884
0.23263
1.07985
1.99283
1.20365
1.90008
1.74322
1.01314
0.96829
0.47431
-0.04933
-0.06528
-0.47413
0.12724
0.70070
1.51743
0.77977
0.84285
0.14275
-0.12590
-0.13327
-0.43116
-0.84361
-0.42295
0.13895
0.48014
0.82428
1.15014
0.54181
0.82265
-0.00205
0.03832
0.16832
-0.33779

١٨٢

49
50
51
52
53
54
55
56
57
58
59
60

49.6
49.9
49.6
50.7
50.7
50.9
50.5
51.2
50.7
50.3
49.2
48.1

50.4979
50.6862
50.7296
50.9166
51.0532
51.1813
51.1869
51.2901
51.2673
51.1350
50.7591
50.1448

50.7224
50.8827
51.0121
50.9708
51.1415
51.2516
51.3586
51.3127
51.4092
51.3438
51.1489
50.6560

-1.12240
-0.98275
-1.41206
-0.27079
-0.44152
-0.35162
-0.85860
-0.11267
-0.70916
-1.04381
-1.94889
-2.55603

Row

Period

Forecast

Lower

Upper

1
2
3
4
5
6

61
62
63
64
65
66

49.8884
49.6319
49.3755
49.1190
48.8626
48.6061

47.5283
47.1597
46.7803
46.3915
45.9946
45.5908

52.2484
52.1041
51.9707
51.8466
51.7306
51.6215

Holt's Double Exponential Smoothing

Actual
Predicted
Forecast

Metals

50

Actual
Predicted
Forecast

45

Smoothing Constants
Alpha (level): 0.200
Gamma (trend):0.300
MAPE:
MAD:
MSD:

40
0

10

20

30

40

50

2.15656
0.96328
1.56274

60

Time

‫ ﻳﺘﻮﻗﻊ ﻋﺪم ﺗﻄﺎﺑﻖ‬:‫ ﺗﺤﻘﻖ ﻣﻦ ﺻﺤﺔ اﻟﺤﺴﺎﺑﺎت اﻟﺴﺎﺑﻘﺔ ﺑﺘﺘﺒﻊ ﺑﻌﺾ اﻟﻘﻴﻢ ﻳﺪوﻳﺎ ) ﻣﻼﺣﻈﺔ‬:‫ﺗﻤﺮﻳﻦ‬
‫اﻟﺤﺴﺎﺑﺎت ﺗﻤﺎﻣﺎ وذﻟﻚ ﻹﺧﺘﻼف ﻃﺮﻳﻖ ﺗﻤﺜﻴﻞ اﻷﻋﺪاد ﺑﻴﻦ اﻟﺤﺎﺳﺐ واﻵﻟﺔ اﻟﺤﺎﺳﺒﺔ وآﺬﻟﻚ ﻓﻲ‬
(‫ﺗﺨﺰﻳﻦ اﻷرﻗﺎم ﻓﻲ ذاآﺮات آﻞ ﻣﻨﻬﻤﺎ‬

١٨٣

‫اﻟﻔﺼﻞ اﻟﺜﺎﻟﺚ ﻋﺸﺮ‬
‫اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺜﻼﺛﻲ و اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﻟﻠﻤﺘﺴﻠﺴﻼت‬
‫اﻟﻤﻮﺳﻤﻴﺔ اﻟﻤﻨﺠﺮﻓﺔ‬
‫‪Triple Exponential Smoothing: Winters' Three‬‬‫‪Parameter Trend and Seasonality Smoothing Method‬‬
‫ﺟﻤﻴﻊ اﻟﻄﺮق اﻟﺴﺎﺑﻘﺔ اﻟﺘﻲ درﺳﻨﺎهﺎ ﻓﻲ هﺬا اﻟﻔﺼﻞ ﻻ ﺗﻨﻔﻊ ﻟﺘﺤﻠﻴﻞ اﻟﻈﻮاهﺮ اﻟﻤﻮﺳﻤﻴﺔ ﻣﺎﻋﺪى‬
‫ﻃﺮﻳﻘﺔ اﻟﺘﻔﻜﻴﻚ ‪ Decomposition Method‬وﻃﺮﻳﻘﺔ وﻧﺘﺮز ‪Winters' trend and‬‬
‫‪ seasonal smoothing‬اﻟﺘﻲ ﺳﻮف ﻧﺴﺘﻌﺮﺿﻬﺎ هﻨﺎ‬

‫ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﻤﻮﺳﻤﻴﺔ اﻟﻤﻨﺠﺮﻓﺔ‬
‫أوﻻ‪ :‬ﺗﻤﻬﺪ اﻟﻤﺸﺎهﺪات ﺗﻤﻬﻴﺪا آﻠﻴﺎ ﺑﺎﻟﻌﻼﻗﺔ‬
‫‪zt‬‬
‫‪+ (1 − α )( st −1 + bt −1 ) , t = 1, 2,..., n‬‬
‫‪St − s‬‬
‫ﺛﺎﻧﻴﺎ‪ :‬ﺗﻤﻬﻴﺪ اﻹﻧﺠﺮاف‬

‫‪st = α‬‬

‫‪bt = γ ( st − st −1 ) + (1 − γ ) bt −1 , t = 1, 2,..., n‬‬

‫ﺛﺎﻟﺜﺎ‪ :‬ﺗﻤﻬﻴﺪ اﻟﻤﻮﺳﻤﻴﺔ‬
‫‪zt‬‬
‫‪+ (1 − β ) St − s , t = 1, 2,..., n‬‬
‫‪st‬‬
‫ﺣﻴﺚ ‪ Si‬هﻲ اﻟﻤﺮآﺒﺔ اﻟﻤﻮﺳﻤﻴﺔ ﻋﻨﺪ اﻟﺰﻣﻦ ‪ i‬و ‪ s‬هﻲ دورة اﻟﻤﻮﺳﻤﻴﺔ‬
‫اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﺗﻌﻄﻲ ﺑﺎﻟﻌﻼﻗﺔ‬
‫‪zˆt = ( st + bt t ) St − s , t = 1, 2,..., n‬‬
‫واﻟﺘﻨﺒﺆات ﻣﻦ اﻟﻌﻼﻗﺔ‬
‫‪zn ( l ) = ( sn + bn l ) Sn − s+l , l > 0‬‬

‫‪St = β‬‬

‫ﻣﻦ اﻟﺼﻌﺐ ﺟﺪا ﺗﺘﺒﻊ ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﺑﺎﻟﺤﺴﺎﺑﺎت اﻟﻴﺪوﻳﺔ ﺣﻴﺚ ان اﻟﻘﻴﻢ اﻻوﻟﻴﺔ ﺗﺤﺴﺐ ﺑﺨﻮارزﻣﺎت‬
‫ﻏﻴﺮ ﺧﻄﻴﺔ ﺑﺈﺳﺘﺨﺪام اﻟﺤﺎﺳﺐ وﻟﻬﺬا ﻟﻦ ﻧﺴﺘﻌﺮﺿﻬﺎ هﻨﺎ وﻧﻜﺘﻔﻲ ﺑﺎﻟﻨﺘﺎﺋﺞ اﻟﻤﺨﺮﺟﺔ ﻣﻦ اﻟﺤﺎﺳﺐ‪.‬‬
‫ﻣﺜﺎل‪:‬‬
‫ﺗﺤﻤﻞ اﻟﺒﻴﺎﻧﺎت ﻣﻦ ورﻗﺔ اﻟﻌﻤﻞ ‪EMPLOY.MTW‬‬
‫‪'E:\Mtbwin\DATA\EMPLOY.MTW'.‬‬

‫‪MTB > Retrieve‬‬

‫ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺸﺎهﺪات ﻓﻲ اﻟﺘﻐﻴﺮ ‪Food‬‬
‫‪Food‬‬

‫‪52.1‬‬
‫‪51.5‬‬
‫‪53.5‬‬
‫‪54.3‬‬

‫‪53.4‬‬
‫‪52.3‬‬
‫‪53.1‬‬
‫‪54.6‬‬

‫‪55.3‬‬
‫‪53.6‬‬
‫‪53.3‬‬
‫‪54.2‬‬

‫‪58.2‬‬
‫‪55.3‬‬
‫‪53.9‬‬
‫‪54.8‬‬

‫‪66.9‬‬
‫‪58.5‬‬
‫‪55.6‬‬
‫‪55.8‬‬

‫‪70.7‬‬
‫‪69.3‬‬
‫‪60.1‬‬
‫‪57.9‬‬
‫‪57.7‬‬

‫‪65.3‬‬
‫‪69.6‬‬
‫‪68.9‬‬
‫‪62.6‬‬
‫‪60.5‬‬

‫‪١٨٤‬‬

‫‪56.5‬‬
‫‪64.2‬‬
‫‪68.8‬‬
‫‪70.3‬‬
‫‪66.4‬‬

‫‪53.4‬‬
‫‪55.5‬‬
‫‪63.6‬‬
‫‪69.4‬‬
‫‪75.5‬‬

‫‪52.5‬‬
‫‪53.3‬‬
‫‪57.1‬‬
‫‪64.7‬‬
‫‪73.3‬‬

‫‪53.2‬‬
‫‪52.4‬‬
‫‪52.2‬‬
‫‪57.1‬‬
‫‪68.1‬‬

‫‪53.0‬‬
‫‪51.5‬‬
‫‪51.5‬‬
‫‪53.9‬‬
‫‪58.1‬‬

‫‪53.5‬‬
‫‪51.5‬‬
‫‪51.7‬‬
‫‪53.5‬‬
‫‪54.8‬‬

‫وﻧﺮﺳﻢ اﻟﻤﺸﺎهﺪات‬

75

Food

70
65
60
55
50
Index

10

20

30

40

50

60

:‫ ﻧﻄﺒﻖ اﻵن ﻃﺮﻳﻘﺔ وﻧﺘﺮز آﺎﻟﺘﺎﻟﻲ‬١٢ ‫ﻧﻼﺣﻆ ان اﻟﻈﺎهﺮة ﻣﻮﺳﻤﻴﺔ ﺑﺪورة‬
Additive Model ‫ ﻟﻠﻨﻤﻮذج اﻹﺿﺎﻓﻲ‬:‫أوﻻ‬
zt = bt + St + et , t = 1, 2,..., n
MTB > %Wintadd 'Food' 12;
SUBC>
Weight 0.2 0.2 0.2;
SUBC>
Forecasts 12;
SUBC>
Title "Wintrs' Trend and Seasonal Smoothing";
SUBC>
Table.

Winters' additive model
Data
Length
NMissing

Food
60.0000
0

Smoothing Constants
Alpha (level):
0.2
Gamma (trend):
0.2
Delta (seasonal): 0.2
Accuracy Measures
MAPE: 1.94769
MAD: 1.15100
MSD: 2.66711

١٨٥

Time
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59

Food
53.5
53.0
53.2
52.5
53.4
56.5
65.3
70.7
66.9
58.2
55.3
53.4
52.1
51.5
51.5
52.4
53.3
55.5
64.2
69.6
69.3
58.5
55.3
53.6
52.3
51.5
51.7
51.5
52.2
57.1
63.6
68.8
68.9
60.1
55.6
53.9
53.3
53.1
53.5
53.5
53.9
57.1
64.7
69.4
70.3
62.6
57.9
55.8
54.8
54.2
54.6
54.3
54.8
58.1
68.1
73.3
75.5
66.4
60.5

Smooth
48.7755
49.6020
51.0736
52.0733
53.5117
57.4851
66.2299
71.7852
71.8932
62.3206
57.5208
55.1544
55.0393
53.6493
53.1185
52.2661
52.6528
55.7616
63.8483
68.8787
68.0386
59.2825
55.0979
53.0432
53.0377
52.1394
51.9889
51.7214
52.1875
55.0750
63.7515
68.8427
68.0160
58.9061
55.3220
53.4419
53.3084
52.6717
52.9095
52.9745
53.7952
57.2065
65.2747
70.4039
69.6200
60.5655
57.0392
55.3721
55.2403
54.6681
54.8138
54.7366
55.3001
58.5310
66.4168
71.8806
71.8794
63.7240
60.3141

Predict
49.4303
50.4197
51.9944
53.0424
54.4591
58.3901
67.0593
72.5443
72.5785
62.7787
57.7958
55.3296
55.1373
53.6258
53.0100
52.0971
52.4960
55.6369
63.7181
68.7678
67.9610
59.2585
55.0435
52.9991
53.0177
52.0907
51.9165
51.6403
52.1009
54.9923
63.7531
68.8382
68.0099
58.9356
55.3981
53.5262
53.4076
52.7666
53.0177
53.1020
53.9386
57.3484
65.4066
70.5076
69.6794
60.6497
57.2014
55.5623
55.4399
54.8422
54.9622
54.8705
55.4112
58.6176
66.4827
72.0112
72.0616
64.0437
60.7281

Error
4.06965
2.58027
1.20556
-0.54244
-1.05914
-1.89013
-1.75932
-1.84430
-5.67851
-4.57874
-2.49577
-1.92957
-3.03734
-2.12584
-1.50996
0.30287
0.80401
-0.13695
0.48187
0.83218
1.33902
-0.75851
0.25650
0.60092
-0.71765
-0.59067
-0.21651
-0.14031
0.09913
2.10774
-0.15314
-0.03822
0.89007
1.16436
0.20190
0.37384
-0.10757
0.33345
0.48233
0.39803
-0.03858
-0.24838
-0.70661
-1.10758
0.62065
1.95031
0.69858
0.23773
-0.63993
-0.64220
-0.36218
-0.57048
-0.61119
-0.51765
1.61731
1.28878
3.43843
2.35629
-0.22810

١٨٦

60

57.7

58.6397

59.0446

-1.34455

Row

Period

Forecast

Lower

Upper

1
2
3
4
5
6
7
8
9
10
11
12

61
62
63
64
65
66
67
68
69
70
71
72

58.6167
58.3236
58.8195
58.9840
59.8723
63.4804
72.0757
77.4486
77.7540
68.9067
64.6434
62.7731

55.7968
55.4449
55.8775
55.9746
56.7913
60.3243
68.8410
74.1321
74.3528
65.4180
61.0647
59.1020

61.4366
61.2023
61.7614
61.9935
62.9532
66.6365
75.3104
80.7651
81.1552
72.3954
68.2221
66.4441

Wintrs' Trend and Seasonal Smoothing

Actual

80

Predicted
Forecast
Actual
Predicted
Forecast

Food

70

60

Smoothing Constants
Alpha (level): 0.200
Gamma (trend):0.200
Delta (season):0.200

50

MAPE:
MAD:
MSD:

0

10

20

30

40

50

60

1.94769
1.15100
2.66711

70

Time

zt = bt St + et , t = 1, 2,..., n

Multiplicative Model ‫ ﻟﻠﻨﻤﻮذج اﻟﺘﻀﺎﻋﻔﻲ‬:‫ﺛﺎﻧﻴﺎ‬

MTB > %Wintmult 'Food' 12;
SUBC>
Weight 0.2 0.2 0.2;
SUBC>
Forecasts 12;
SUBC>
Title "Winters' Trend and Seasonal Smoothing";
SUBC>
Table.

Winters' multiplicative model
Data

Food

١٨٧

Length
NMissing

60.0000
0

Smoothing Constants
Alpha (level):
0.2
Gamma (trend):
0.2
Delta (seasonal): 0.2
Accuracy Measures
MAPE: 1.88377
MAD: 1.12068
MSD: 2.86696

Time
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44

Food
53.5
53.0
53.2
52.5
53.4
56.5
65.3
70.7
66.9
58.2
55.3
53.4
52.1
51.5
51.5
52.4
53.3
55.5
64.2
69.6
69.3
58.5
55.3
53.6
52.3
51.5
51.7
51.5
52.2
57.1
63.6
68.8
68.9
60.1
55.6
53.9
53.3
53.1
53.5
53.5
53.9
57.1
64.7
69.4

Smooth

48.7870
49.6755
51.1521
52.1675
53.6181
57.6509
66.6199
72.4105
72.5679
62.7837
57.9154
55.5108
54.4920
53.2117
52.8118
52.0929
52.5894
55.7388
63.7189
68.7087
67.9722
59.4594
55.4037
53.4103
52.6818
51.8659
51.8002
51.6271
52.1643
55.0424
63.6079
68.6702
67.9561
59.1021
55.6210
53.7881
53.0479
52.4502
52.7444
52.8747
53.7689
57.2790
65.4702
70.6713

Predict

49.3853
50.4303
52.0132
53.0746
54.5132
58.5541
67.5607
73.3280
73.3777
63.2634
58.1732
55.6485
54.5392
53.1621
52.6957
51.9302
52.4439
55.6209
63.5782
68.5838
67.8890
59.4361
55.3468
53.3536
52.6356
51.8071
51.7290
51.5549
52.0890
54.9676
63.6199
68.6823
67.9727
59.1487
55.7003
53.8609
53.1211
52.5294
52.8467
53.0029
53.9188
57.4374
65.6357
70.8095

Error

4.11470
2.56966
1.18677
-0.57458
-1.11323
-2.05414
-2.26072
-2.62800
-6.47768
-5.06337
-2.87320
-2.24849
-2.43920
-1.66212
-1.19573
0.46985
0.85611
-0.12087
0.62178
1.01617
1.41104
-0.93606
-0.04680
0.24639
-0.33562
-0.30705
-0.02902
-0.05492
0.11103
2.13244
-0.01988
0.11774
0.92727
0.95133
-0.10032
0.03912
0.17892
0.57055
0.65329
0.49714
-0.01879
-0.33743
-0.93567
-1.40954

١٨٨

45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

70.3
62.6
57.9
55.8
54.8
54.2
54.6
54.3
54.8
58.1
68.1
73.3
75.5
66.4
60.5
57.7

69.8908
60.7552
57.1925
55.5181
54.8764
54.3244
54.5372
54.5298
55.1925
58.6054
66.7739
72.4056
72.3385
63.6729
60.0395
58.3023

69.9719
60.8370
57.3348
55.6775
55.0383
54.4749
54.6769
54.6661
55.3155
58.7141
66.8698
72.5622
72.5236
63.9378
60.3781
58.6338

0.32815
1.76302
0.56523
0.12253
-0.23826
-0.27486
-0.07694
-0.36612
-0.51551
-0.61410
1.23016
0.73784
2.97638
2.46217
0.12191
-0.93381

Row

Period

Forecast

Lower

Upper

1
2
3
4
5
6
7
8
9
10
11
12

61
62
63
64
65
66
67
68
69
70
71
72

57.8102
57.3892
57.8332
57.9307
58.8311
62.7415
72.1849
78.1507
78.5092
68.6689
63.9258
61.8189

55.0645
54.5864
54.9687
55.0005
55.8313
59.6686
69.0354
74.9215
75.1976
65.2721
60.4414
58.2446

60.5558
60.1921
60.6977
60.8609
61.8309
65.8145
75.3344
81.3798
81.8208
72.0657
67.4103
65.3933

Winters' Trend and Seasonal Smoothing

Actual

80

Predicted
Forecast
Actual
Predicted
Forecast

Food

70

60

Smoothing Constants
Alpha (level): 0.200
Gamma (trend):0.200
Delta (season):0.200

50

MAPE:
MAD:
MSD:

0

10

20

30

40

50

60

1.88377
1.12068
2.86696

70

Time

:‫ﻣﻼﺣﻈﺎت‬
‫ ﺛﺎﺑﺖ‬γ ‫ ﺛﺎﺑﺖ اﻟﺘﻤﻬﻴﺪ اﻟﻜﻠﻲ و‬α ‫ﻟﺘﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﻧﺤﺘﺎج اﻟﻲ إﺧﺘﻴﺎر ﻗﻴﻢ ﺛﻼﺛﺔ ﻣﻌﺎﻟﻢ هﻲ‬
‫ ﻓﻲ ﺛﻼﺛﺔ‬Optimization ‫ ﺛﺎﺑﺖ ﺗﻤﻬﻴﺪ اﻟﻤﻮﺳﻤﻴﺔ وهﺬﻩ ﻋﻤﻠﻴﺔ أﻓﻀﻠﻴﺔ‬β ‫ﺗﻤﻬﻴﺪ اﻹﻧﺠﺮاف و‬
١٨٩

‫اﺑﻌﺎد ) ﻓﻀﺎء اﻟﻤﻌﺎﻟﻢ ( إﻣﺎ أن ﻧﺘﺮك ﻟﻠﺒﺮﻧﺎﻣﺞ اﻹﺣﺼﺎﺋﻲ اﻟﻤﺴﺘﺨﺪم ﺣﺴﺎﺑﻬﺎ ﺗﻠﻘﺎﺋﻴﺎ ﺑﺈﺳﺘﺨﺪام‬
‫ﺧﻮارزﻣﺎت ﻏﻴﺮ ﺧﻄﻴﺔ ﻣﺒﻨﻴﺔ داﺧﻞ اﻟﺒﺮﻧﺎﻣﺞ أو ﻧﻘﻮم ﻧﺤﻦ ﺑﺈﻣﺪاد اﻟﺒﺮﻧﺎﻣﺞ ﺑﺘﻠﻚ اﻟﻘﻴﻢ‪.‬‬
‫ﺗﻤﺮﻳﻦ‪ :‬ﻓﻲ اﻷﻣﺜﻠﺔ اﻟﺴﺎﺑﻘﺔ اﺧﺬﻧﺎ ‪ . α = γ = β = 0.2‬ﺛﺒﺖ ﻓﻲ آﻞ ﻣﺮة ﻣﻌﻠﻤﻴﻦ وﻏﻴﺮ اﻟﺜﺎﻟﺚ‬
‫ﺣﺘﻲ ﺗﺤﺼﻞ ﻋﻠﻲ أﻗﻞ ‪MSD‬‬
‫ﻣﺜﺎل ﻋﻤﻠﻲ ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ‪:‬‬
‫ﺳﻮف ﻧﺴﺘﺨﺪم ورﻗﺔ اﻟﻌﻤﻞ ‪ CPI.MTW‬ﻣﻦ ﻣﺠﻤﻮﻋﺔ اﻟﺒﻴﺎﻧﺎت ﻟﻠﺒﺮﻧﺎﻣﺞ ‪MINITAB‬‬
‫‪MTB > Retrieve‬‬

‫‪'C:\MTBWIN\STUDENT9\CPI.MTW'.‬‬

‫ﺳﻮف ﻧﺄﺧﺬ اﻟﻤﺘﻐﻴﺮ ‪CPIChange‬‬
‫‪2.9‬‬
‫‪6.2‬‬
‫‪13.5‬‬
‫‪3.6‬‬

‫‪1.6‬‬
‫‪3.2‬‬
‫‪11.3‬‬
‫‪1.9‬‬

‫‪1.3‬‬
‫‪5.7‬‬
‫‪6.5‬‬
‫‪4.3‬‬
‫‪4.2‬‬

‫‪1.3‬‬
‫‪4.4‬‬
‫‪7.6‬‬
‫‪3.6‬‬
‫‪3.0‬‬

‫‪1.0‬‬
‫‪4.2‬‬
‫‪9.1‬‬
‫‪6.2‬‬
‫‪4.8‬‬

‫‪1.0‬‬
‫‪5.5‬‬
‫‪5.8‬‬
‫‪3.2‬‬
‫‪5.4‬‬

‫‪CPIChnge‬‬
‫‪1.7‬‬
‫‪3.1‬‬
‫‪11.0‬‬
‫‪10.3‬‬
‫‪4.1‬‬

‫ﻧﺮﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ وﻧﻮﺟﺪ اﻟﺘﺮﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ اﻟﻌﻴﻨﻴﺔ‬
‫‪MTB > TSPlot‬‬

‫‪14‬‬
‫‪12‬‬
‫‪10‬‬

‫‪6‬‬

‫‪CPIChnge‬‬

‫‪8‬‬

‫‪4‬‬
‫‪2‬‬
‫‪0‬‬
‫‪30‬‬

‫‪20‬‬

‫‪25‬‬

‫‪10‬‬

‫‪15‬‬

‫‪Index‬‬

‫‪5‬‬

‫‪MTB > %acf c2‬‬
‫‪Autocorrelation Function for CPIChnge‬‬

‫‪8‬‬

‫‪7‬‬

‫‪5‬‬

‫‪6‬‬

‫‪4‬‬

‫‪Corr‬‬

‫‪Lag‬‬

‫‪LBQ‬‬

‫‪T‬‬

‫‪41.23‬‬

‫‪8 -0.16 -0.51‬‬

‫‪١٩٠‬‬

‫‪3‬‬

‫‪2‬‬

‫‪Corr‬‬

‫‪1‬‬

‫‪Lag‬‬

‫‪LBQ‬‬

‫‪T‬‬

‫‪22.42‬‬
‫‪30.32‬‬
‫‪33.59‬‬
‫‪36.24‬‬
‫‪38.97‬‬
‫‪40.05‬‬
‫‪40.07‬‬

‫‪1 0.79 4.53‬‬
‫‪2 0.46 1.77‬‬
‫‪3 0.29 1.02‬‬
‫‪4 0.26 0.88‬‬
‫‪5 0.26 0.86‬‬
‫‪6 0.16 0.52‬‬
‫‪7 -0.02 -0.06‬‬

‫‪Autocorrelation‬‬

‫‪1.0‬‬
‫‪0.8‬‬
‫‪0.6‬‬
‫‪0.4‬‬
‫‪0.2‬‬
‫‪0.0‬‬
‫‪-0.2‬‬
‫‪-0.4‬‬
‫‪-0.6‬‬
‫‪-0.8‬‬
‫‪-1.0‬‬

MTB > %pacf c2
Partial Autocorrelation

Partial Autocorrelation Function for CPIChnge
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

1

2

3

4

Lag PAC
1
2
3
4
5
6
7

0.79
-0.42
0.35
-0.04
0.08
-0.28
-0.05

5

6

T

Lag PAC

T

4.53
-2.44
2.01
-0.24
0.43
-1.62
-0.29

8 -0.09

-0.52

7

8

ARMA(١،١) ‫ﻣﻦ اﻧﻤﺎط اﻟﺘﺮﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ اﻟﻌﻴﻨﻴﺔ ﻗﺪ ﻳﻜﻮن اﻟﻨﻤﻮذج‬
‫ ﺗﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬٥ ‫ اﻷواﻣﺮ اﻟﺘﺎﻟﻴﺔ ﺗﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح وﺗﻮﻟﺪ‬،‫ﻳﻨﻄﺒﻖ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ‬
MTB >
SUBC>
SUBC>
SUBC>
SUBC>
SUBC>
SUBC>

arima 1 0 1 c2;
fore 5 c3 c4 c5;
gser;
gacf;
gpacf;
ghist;
gnormal.

ARIMA Model
ARIMA model for CPIChnge
Estimates at each iteration
Iteration
SSE
Parameters
0
323.251
0.100
0.100
4.522
1
200.616
0.250
-0.050
3.745
2
182.146
0.184
-0.200
4.067
3
163.067
0.135
-0.350
4.308
4
142.864
0.107
-0.500
4.434
5
121.402
0.111
-0.650
4.407
6
99.668
0.150
-0.800
4.197
7
77.036
0.268
-0.950
3.590
8
67.550
0.418
-0.956
2.828
9
62.802
0.568
-0.964
2.062
10
62.108
0.637
-0.973
1.687
11
62.030
0.644
-0.979
1.619
12
62.003
0.647
-0.982
1.584
13
61.996
0.651
-0.985
1.549
14
61.996
0.651
-0.986
1.539
Unable to reduce sum of squares any further
Final Estimates of Parameters

١٩١

Type
AR
1
MA
1
Constant
Mean

Coef
0.6513
-0.9857
1.5385
4.412

StDev
0.1434
0.0516
0.4894
1.403

T
4.54
-19.11
3.14

Number of observations: 33
Residuals:
SS = 61.8375 (backforecasts excluded)
MS = 2.0613 DF = 30
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
9.6(DF=10)
17.0(DF=22)
* (DF= *)

48
* (DF= *)

Forecasts from period 33
Period
34
35
36
37
38

95 Percent Limits
Lower
Upper
0.3216
5.9507
-1.8180
8.9801
-2.3061
10.0476
-2.4192
10.5380
-2.4201
10.7848

Forecast
3.1362
3.5810
3.8708
4.0594
4.1823

zt = 1.54 + 0.65zt −1 + at − 0.99at −1 , at

Actual

‫أي ان اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬

N ( 0, 2.06 )

‫ هﻲ‬t ‫ﻣﻘﺪرات اﻟﻤﻌﺎﻟﻢ وإﻧﺤﺮاﻓﺎﺗﻬﺎ اﻟﻤﻌﻴﺎرﻳﺔ وﻗﻴﻤﺔ إﺧﺘﺒﺎر‬
ˆ
ˆ
φ1 = 0.6513, s.e. φ1 = 0.1434, t = 4.54

( )
θˆ = −0.9857, s.e. (θˆ ) = 0.0516, t = −19.11
δˆ = 1.5385, s.e. (δˆ ) = 0.4894, t = 3.14
1

1

σˆ 2 = 2.0613, with d . f . = 30

.‫ﻧﻼﺣﻆ ان ﺟﻤﻴﻊ اﻟﻤﻌﺎﻟﻢ ﻣﻌﻨﻮﻳﺔ‬

:‫اﻵن ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬

ACF of Residuals for CPIChnge
(with 95% confidence limits for the autocorrelations)
1.0
0.8

Autocorrelation

0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

١٩٢

PACF of Residuals for CPIChnge
(with 95% confidence limits for the partial autocorrelations)
1.0

Partial Autocorrelation

0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
1

2

3

4

5

6

7

8

Lag

‫أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ ﺗﺪل ﻋﻠﻰ أن اﻟﺒﻮاﻗﻲ ﺗﺘﺒﻊ ﺗﻮزﻳﻊ ﺿﺠﺔ‬
:‫ ﻟﻨﻔﺤﺺ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬،‫ﺑﻴﻀﺎء أي ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ‬
‫رﺳﻢ اﻟﻤﺪرج اﻟﺘﻜﺮاري‬

Histogram of the Residuals
(response is CPIChnge)
8
7

Frequency

6
5
4
3
2

١٩٣

.‫ﻳﺒﺪو ﻣﺘﻨﺎﻇﺮ ﺑﻌﺾ اﻟﺸﻴﺊ‬
:‫ﻟﻨﻨﻈﺮ إﻟﻰ ﻣﺨﻄﻂ اﻹﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ ﻟﻠﺒﻮاﻗﻲ‬

Normal Probability Plot of the Residuals
(response is CPIChnge)
4
3

Residual

2
1
0
-1
-2
-3
-2

-1

0

1

2

Normal Score

.‫ﻧﺴﺘﻄﻴﻊ أن ﻧﻘﻮل ان اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ ﺗﻘﺮﻳﺒﺎ‬
.‫ ﺗﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬٥ ‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﻤﺘﺴﻠﺴﺔ ﻣﻊ‬

Time Series Plot for CPIChnge
(with forecasts and their 95% confidence limits)

CPIChnge

10

5

0

١٩٤

:‫ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ آﺎﻟﺘﺎﻟﻲ‬AR(٢) ‫دﻋﻨﺎ ﻧﺤﺎول ﺗﻄﺒﻴﻖ ﻧﻤﻮذج‬
MTB > arima 2 0 0 c2
Type
AR
1
AR
2
Constant
Mean

Coef
1.1872
-0.4657
1.3270
4.765

StDev
0.1625
0.1624
0.2996
1.076

T
7.31
-2.87
4.43

Number of observations: 33
Residuals:
SS = 88.6206 (backforecasts excluded)
MS = 2.9540 DF = 30
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
19.8(DF=10)
25.4(DF=22)
* (DF= *)

zt = 1.33 + 1.187 zt −1 − 0.4657 zt −1 + at , at

N ( 0, 2.95)

48
* (DF= *)

‫أي ان اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬

‫ هﻲ‬t ‫ﻣﻘﺪرات اﻟﻤﻌﺎﻟﻢ وإﻧﺤﺮاﻓﺎﺗﻬﺎ اﻟﻤﻌﻴﺎرﻳﺔ وﻗﻴﻤﺔ إﺧﺘﺒﺎر‬

( )
φˆ = −0.4657, s.e. (θˆ ) = 0.1624, t = −2.87
δˆ = 1.327, s.e. (δˆ ) = 0.2996, t = 4.43

φˆ1 = 1.1872, s.e. φˆ1 = 0.1625, t = 7.31
2

1

σˆ 2 = 2.954, with d . f . = 30

‫ﻟﻨﻨﻈﺮ إﻟﻰ اﻹﺧﺘﺒﺎر‬

H 0 : φ2 = 0
H 1 : φ2 ± 0

‫اﻹﺣﺼﺎﺋﺔ‬
t0 =

φˆ

2

( )

s.e. φˆ2

=

−0.4657
= −2.8676
0.1624

‫ ﻟﻬﺎ ﺑﺎﻷﻣﺮ‬P-value ‫ﻧﻮﺟﺪ اﻟـ‬
MTB > cdf -2.8676;
SUBC> t 30.

Cumulative Distribution Function

١٩٥

‫‪Student's t distribution with 30 DF‬‬
‫)‪P( X <= x‬‬
‫‪0.0037‬‬

‫‪x‬‬
‫‪-2.8676‬‬

‫أي اﻟـ ‪ P-value‬ﻟﻬﺎ ﺗﺴﺎوي ‪ ٠٫٠٠٣٧‬وهﻲ أﻗﻞ ﻣﻦ ‪ ٠٫٠٥‬أي ﻻﻧﺮﻓﺾ ان ‪ φ2 = 0‬وﺑﺎﻟﺘﺎﻟﻲ‬
‫ﻧﺮﻓﺾ اﻟﻨﻤﻮذج )‪. AR(٢‬‬
‫ﺗﻤﺮﻳﻦ‪:‬‬
‫ﺣﺎول ﺗﻄﺒﻴﻖ ﻧﻤﺎذج اﺧﺮى ﻣﻨﺎﺳﺒﺔ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺴﺎﺑﻘﺔ وإﺧﺘﺎر أﻓﻀﻞ ﻧﻤﻮذج‪ ،‬أﺟﺮي‬
‫اﻹﺧﺘﺒﺎرات اﻟﻤﻨﺎﺳﺒﺔ واﺳﺘﺨﺪم اﻳﻀﺎ اﻟﻤﻌﻴﺎر ‪. AIC‬‬
‫ﻣﺜﺎل ﻋﻤﻠﻲ ﺁﺧﺮ ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ‪:‬‬
‫ﺳﻮف ﻧﺤﺎول ﺑﻨﺎء ﻧﻤﻮذج ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬
‫)‪z(t‬‬
‫‪-103.2‬‬
‫‪-391.4‬‬
‫‪-836.1‬‬
‫‪-1307.3‬‬
‫‪-1736.9‬‬
‫‪-2097.5‬‬
‫‪-2363.4‬‬
‫‪-2721.4‬‬
‫‪-3353.3‬‬
‫‪-4153.0‬‬
‫‪-5068.7‬‬
‫‪-5848.8‬‬
‫‪-6387.3‬‬
‫‪-6834.5‬‬
‫‪-7259.9‬‬
‫‪-7891.4‬‬
‫‪-8791.2‬‬
‫‪-9878.4‬‬
‫‪-11071.9‬‬
‫‪-12157.5‬‬
‫‪-13190.3‬‬
‫‪-14196.3‬‬
‫‪-15335.9‬‬
‫‪-16697.1‬‬
‫‪-18155.6‬‬
‫‪-19547.0‬‬
‫‪-20827.7‬‬
‫‪-21908.3‬‬

‫‪-76.7‬‬
‫‪-339.9‬‬
‫‪-766.7‬‬
‫‪-1242.9‬‬
‫‪-1679.0‬‬
‫‪-2055.4‬‬
‫‪-2328.5‬‬
‫‪-2651.9‬‬
‫‪-3253.4‬‬
‫‪-4028.2‬‬
‫‪-4937.2‬‬
‫‪-5760.6‬‬
‫‪-6317.8‬‬
‫‪-6773.3‬‬
‫‪-7193.3‬‬
‫‪-7784.7‬‬
‫‪-8649.8‬‬
‫‪-9713.3‬‬
‫‪-10903.2‬‬
‫‪-12011.7‬‬
‫‪-13039.8‬‬
‫‪-14051.7‬‬
‫‪-15158.3‬‬
‫‪-16492.7‬‬
‫‪-17951.5‬‬
‫‪-19356.3‬‬
‫‪-20655.7‬‬
‫‪-21762.1‬‬

‫‪-52.8‬‬
‫‪-291.3‬‬
‫‪-698.2‬‬
‫‪-1177.4‬‬
‫‪-1620.8‬‬
‫‪-2010.0‬‬
‫‪-2290.6‬‬
‫‪-2589.5‬‬
‫‪-3156.1‬‬
‫‪-3906.6‬‬
‫‪-4805.0‬‬
‫‪-5663.8‬‬
‫‪-6244.3‬‬
‫‪-6711.8‬‬
‫‪-7131.7‬‬
‫‪-7683.2‬‬
‫‪-8512.3‬‬
‫‪-9552.0‬‬
‫‪-10734.0‬‬
‫‪-11863.9‬‬
‫‪-12889.7‬‬
‫‪-13910.1‬‬
‫‪-14986.4‬‬
‫‪-16290.9‬‬
‫‪-17745.2‬‬
‫‪-19161.7‬‬
‫‪-20478.0‬‬
‫‪-21614.0‬‬

‫‪-33.1‬‬
‫‪-246.0‬‬
‫‪-631.6‬‬
‫‪-1111.3‬‬
‫‪-1561.5‬‬
‫‪-1960.6‬‬
‫‪-2250.5‬‬
‫‪-2533.0‬‬
‫‪-3060.8‬‬
‫‪-3788.7‬‬
‫‪-4673.3‬‬
‫‪-5558.0‬‬
‫‪-6167.9‬‬
‫‪-6649.0‬‬
‫‪-7073.1‬‬
‫‪-7586.4‬‬
‫‪-8379.1‬‬
‫‪-9394.6‬‬
‫‪-10564.0‬‬
‫‪-11713.8‬‬
‫‪-12740.8‬‬
‫‪-13769.9‬‬
‫‪-14819.8‬‬
‫‪-16091.5‬‬
‫‪-17535.7‬‬
‫‪-18965.1‬‬
‫‪-20294.5‬‬
‫‪-21463.7‬‬
‫‪-22474.4‬‬

‫وﻟﻬﺎ اﻟﺮﺳﻢ اﻟﺰﻣﻨﻲ اﻟﺘﺎﻟﻲ‪:‬‬

‫‪١٩٦‬‬

‫‪-19.2‬‬
‫‪-204.0‬‬
‫‪-566.7‬‬
‫‪-1044.2‬‬
‫‪-1500.0‬‬
‫‪-1906.9‬‬
‫‪-2210.8‬‬
‫‪-2482.0‬‬
‫‪-2968.6‬‬
‫‪-3675.0‬‬
‫‪-4542.3‬‬
‫‪-5444.4‬‬
‫‪-6090.4‬‬
‫‪-6584.6‬‬
‫‪-7015.6‬‬
‫‪-7495.3‬‬
‫‪-8249.2‬‬
‫‪-9239.6‬‬
‫‪-10392.1‬‬
‫‪-11560.2‬‬
‫‪-12593.4‬‬
‫‪-13629.3‬‬
‫‪-14657.5‬‬
‫‪-15895.8‬‬
‫‪-17324.4‬‬
‫‪-18766.2‬‬
‫‪-20107.3‬‬
‫‪-21310.9‬‬
‫‪-22335.1‬‬

‫‪-9.1‬‬
‫‪-165.9‬‬
‫‪-504.8‬‬
‫‪-975.1‬‬
‫‪-1436.5‬‬
‫‪-1851.1‬‬
‫‪-2173.3‬‬
‫‪-2437.6‬‬
‫‪-2880.4‬‬
‫‪-3564.3‬‬
‫‪-4412.0‬‬
‫‪-5323.6‬‬
‫‪-6011.8‬‬
‫‪-6518.9‬‬
‫‪-6957.1‬‬
‫‪-7411.2‬‬
‫‪-8124.0‬‬
‫‪-9086.3‬‬
‫‪-10219.0‬‬
‫‪-11402.0‬‬
‫‪-12447.5‬‬
‫‪-13486.3‬‬
‫‪-14499.7‬‬
‫‪-15705.0‬‬
‫‪-17113.5‬‬
‫‪-18564.2‬‬
‫‪-19919.5‬‬
‫‪-21154.4‬‬
‫‪-22195.6‬‬

‫‪-2.5‬‬
‫‪-132.4‬‬
‫‪-446.4‬‬
‫‪-905.3‬‬
‫‪-1371.8‬‬
‫‪-1794.5‬‬
‫‪-2136.4‬‬
‫‪-2398.7‬‬
‫‪-2797.6‬‬
‫‪-3456.9‬‬
‫‪-4281.7‬‬
‫‪-5197.8‬‬
‫‪-5931.6‬‬
‫‪-6453.5‬‬
‫‪-6896.0‬‬
‫‪-7332.9‬‬
‫‪-8004.6‬‬
‫‪-8936.4‬‬
‫‪-10047.2‬‬
‫‪-11238.9‬‬
‫‪-12302.5‬‬
‫‪-13339.6‬‬
‫‪-14345.9‬‬
‫‪-15518.4‬‬
‫‪-16904.2‬‬
‫‪-18360.1‬‬
‫‪-19733.5‬‬
‫‪-20993.7‬‬
‫‪-22053.2‬‬

O r ig in a l T im e S e r ie s

-1 0 0 0 0

-2 0 0 0 0

In d e x

50

100

150

200

:‫اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ هﻲ‬

Autocorrelation Function for z(t)
Autocorrelation

1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

10
Lag Corr
1
2
3
4
5
6
7
8
9
10
11
12

T

LBQ

0.9813.93 196.91
0.97 8.00 388.73
0.95 6.15 575.44
0.94 5.15 757.06
0.92 4.50 933.59
0.91 4.041105.06
0.89 3.681271.47
0.88 3.391432.85
0.86 3.151589.23
0.84 2.951740.66
0.83 2.781887.17
0.81 2.622028.81

Lag Corr
13
14
15
16
17
18
19
20
21
22
23
24

20
T

LBQ

2.49
2165.64
2.37
2297.73
2.25
2425.14
2.15
2547.94
2.06
2666.22
1.97
2780.05
1.89
2889.51
1.82
2994.69
1.75
3095.67
1.68
3192.54
1.62
3285.39
1.56
3374.32

0.80
0.78
0.76
0.75
0.73
0.72
0.70
0.68
0.67
0.65
0.64
0.62

30

Lag Corr
25
26
27
28
29
30
31
32
33
34
35
36

0.61
0.59
0.58
0.56
0.55
0.53
0.52
0.50
0.49
0.47
0.46
0.45

T

LBQ

1.50
3459.42
1.45
3540.78
1.40
3618.50
1.35
3692.67
1.30
3763.38
1.26
3830.74
1.21
3894.84
1.17
3955.76
1.13
4013.61
1.09
4068.47
1.05
4120.44
1.01
4169.59

40

Lag Corr
37
38
39
40
41
42
43
44
45
46
47
48

0.43
0.42
0.41
0.39
0.38
0.36
0.35
0.34
0.32
0.31
0.30
0.28

T

LBQ

50

Lag Corr

T

LBQ

49 0.27 0.594589.24
50 0.26 0.564607.24

0.98
4216.02
0.94
4259.81
0.91
4301.04
0.87
4339.79
0.84
4376.14
0.81
4410.16
0.77
4441.93
0.74
4471.52
0.71
4498.99
0.68
4524.43
0.65
4547.90
0.62
4569.48

P artial A utocorrelation Function for z(t)
Partial Autocorrelation

z(t)

0

1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

10

20

30

40

Lag

PAC

T

Lag

PAC

T

Lag

PAC

T

Lag

PAC

T

1
2
3
4
5
6
7
8
9
10
11
12

0.98
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01

13.93
-0.17
-0.17
-0.16
-0.16
-0.16
-0.15
-0.15
-0.15
-0.14
-0.14
-0.13

13
14
15
16
17
18
19
20
21
22
23
24

-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01

-0.13
-0.12
-0.12
-0.11
-0.11
-0.11
-0.11
-0.10
-0.10
-0.09
-0.09
-0.09

25
26
27
28
29
30
31
32
33
34
35
36

-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01

-0.09
-0.09
-0.09
-0.08
-0.08
-0.08
-0.08
-0.08
-0.08
-0.08
-0.09
-0.09

37
38
39
40
41
42
43
44
45
46
47
48

-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01

-0.09
-0.09
-0.10
-0.10
-0.11
-0.11
-0.12
-0.12
-0.12
-0.13
-0.13
-0.13

50
Lag

PAC

T

49 -0.01
50 -0.01

-0.14
-0.14

.‫ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ‬zt ‫واﺿﺢ ﺟﺪا ان اﻟﻤﺘﺴﻠﺴﻠﺔ‬
‫ وﻧﺮﺳﻤﻬﺎ‬wt = zt − zt −1 ‫ﻧﺄﺧﺬ اﻟﻔﺮوق اﻻوﻟﻰ‬
١٩٧

F irs t D if f e re n c e s w (t)= z (t)-z (t-1 )

w(t)

0

-1 0 0

-2 0 0
In d e x

50

100

150

200

:‫اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻔﺮﻗﺔ هﻲ‬
Autocorrelation

Autocorrelation Function for w(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

5
Lag Corr
1
2
3
4
5
6
7
8
9
10
11
12

T

LBQ

0.9913.93 197.10
0.97 7.98 389.03
0.95 6.12 574.74
0.93 5.10 753.55
0.91 4.43 924.81
0.89 3.941087.93
0.86 3.561242.44
0.83 3.251388.10
0.81 2.981524.83
0.78 2.761652.58
0.75 2.561771.42
0.72 2.381881.49

15

Lag Corr
13
14
15
16
17
18
19
20
21
22
23
24

0.69
0.66
0.63
0.60
0.57
0.54
0.52
0.49
0.47
0.45
0.43
0.41

T

LBQ

2.21
1983.03
2.07
2076.38
1.93
2161.98
1.81
2240.33
1.70
2311.99
1.60
2377.42
1.50
2437.13
1.41
2491.57
1.33
2541.23
1.26
2586.61
1.19
2628.22
1.13
2666.48

25
Lag Corr
25
26
27
28
29
30
31
32
33
34
35
36

0.39
0.38
0.37
0.36
0.35
0.34
0.34
0.34
0.34
0.34
0.34
0.34

T

LBQ

1.08
2701.91
1.03
2734.99
1.00
2766.16
0.96
2795.84
0.94
2824.46
0.92
2852.35
0.91
2879.79
0.90
2907.03
0.89
2934.32
0.89
2961.85
0.89
2989.82
0.89
3018.34

35
Lag Corr
37
38
39
40
41
42
43
44
45
46
47
48

0.34
0.35
0.35
0.35
0.35
0.35
0.35
0.34
0.34
0.33
0.33
0.32

T

45
LBQ

0.90
3047.50
0.90
3077.32
0.90
3107.68
0.90
3138.42
0.90
3169.33
0.89
3200.21
0.88
3230.81
0.87
3260.99
0.85
3290.64
0.84
3319.59
0.82
3347.55
0.79
3374.17

Lag Corr

T

LBQ

49 0.31 0.763399.16

Partial Autocorrelation

P a rtia l A u to c o rre la tio n F u n c tio n fo r w (t)
1 .0
0 .8
0 .6
0 .4
0 .2
0 .0
-0 .2
-0 .4
-0 .6
-0 .8
-1 .0

5
Lag

P AC

T

1
2
3
4
5
6
7
8
9
10
11
12

0 .9 9
-0 .1 4
-0 .1 1
-0 .0 7
-0 .0 6
-0 .0 7
-0 .0 6
-0 .0 3
-0 .0 3
-0 .0 3
-0 .0 2
-0 .0 2

1 3 .9 3
-1 .9 7
-1 .5 1
-0 .9 6
-0 .9 0
-0 .9 7
-0 .7 8
-0 .4 8
-0 .3 7
-0 .3 7
-0 .3 4
-0 .2 2

15

Lag

P AC

T

1 3 -0 .0 1
1 4 -0 .0 0
1 5 0 .0 1
1 6 0 .0 2
1 7 0 .0 2
1 8 0 .0 1
1 9 0 .0 1
2 0 0 .0 1
2 1 0 .0 1
2 2 0 .0 3
2 3 0 .0 2
2 4 0 .0 2

-0 .1 6
-0 .0 2
0 .1 0
0 .2 3
0 .2 5
0 .1 0
0 .1 2
0 .1 1
0 .1 6
0 .3 5
0 .3 1
0 .2 7

25
Lag

35

45

P AC

T

Lag

P AC

T

2 5 0 .0 4
2 6 0 .0 4
2 7 0 .0 3
2 8 0 .0 4
2 9 0 .0 4
3 0 0 .0 2
3 1 0 .0 2
3 2 0 .0 1
3 3 0 .0 1
3 4 0 .0 1
3 5 0 .0 0
3 6 -0 .0 0

0 .5 2
0 .6 2
0 .4 7
0 .5 5
0 .5 6
0 .3 4
0 .2 2
0 .1 6
0 .1 8
0 .1 5
0 .0 4
-0 .0 7

37
38
39
40
41
42
43
44
45
46
47
48

-0 .0 1
-0 .0 2
-0 .0 4
-0 .0 4
-0 .0 4
-0 .0 4
-0 .0 3
-0 .0 0
0 .0 1
-0 .0 1
-0 .0 4
-0 .0 5

-0 .0 7
-0 .2 9
-0 .6 3
-0 .6 0
-0 .5 4
-0 .5 5
-0 .4 2
-0 .0 7
0 .0 8
-0 .1 8
-0 .5 6
-0 .7 6

Lag

P AC

T

4 9 -0 .0 3

-0 .4 8

.‫ ﻻﺗﺰال ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ‬wt ‫واﺿﺢ ﺟﺪا ان اﻟﻤﺘﺴﻠﺴﻠﺔ‬
‫ )ﻻﺣﻆ ان هﺬا اﻟﻔﺮق اﻟﺜﺎﻧﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻷﺻﻠﻴﺔ( وﻧﺮﺳﻤﻬﺎ‬yt = wt − wt −1 ‫ﻧﺄﺧﺬ اﻟﻔﺮوق اﻻوﻟﻰ‬

١٩٨

F ir s t D if f e r e n c e s

y (t)= w (t)-w (t-1 )

1 0

y(t)

5

0

-5

In d e x

5 0

1 0 0

1 5 0

2 0 0

:‫اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ ﻟﻬﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻔﺮﻗﺔ هﻲ‬
Autocorrelation

A utocorrelation Function for y(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

5
Lag Corr
1
2
3
4
5
6
7
8
9
10
11
12

T

LBQ

0.9313.08 173.62
0.80 6.83 303.36
0.70 4.89 401.81
0.62 3.89 479.62
0.56 3.26 542.93
0.50 2.77 593.77
0.43 2.32 632.35
0.36 1.90 659.84
0.29 1.50 677.66
0.22 1.11 687.71
0.15 0.75 692.41
0.08 0.41 693.82

15

Lag Corr

T

LBQ

25
Lag Corr

13 0.01 0.06 693.85
14 -0.05 -0.23 694.31
15 -0.09 -0.46 696.12
16 -0.14 -0.70 700.34
17 -0.19 -0.97 708.51
18 -0.25 -1.24 722.16
19 -0.31 -1.53 743.30
20 -0.37 -1.83 774.40
21 -0.44 -2.13 818.29
22 -0.51 -2.40 876.73
23 -0.56 -2.57 948.00
1030.79
24 -0.60 -2.67

T

LBQ

1124.21
25 -0.64 -2.73
1224.05
26 -0.66 -2.72
1324.28
27 -0.66 -2.62
1418.06
28 -0.63 -2.44
1500.63
29 -0.59 -2.22
1572.45
30 -0.55 -2.01
1632.42
31 -0.50 -1.80
1677.44
32 -0.43 -1.53
1708.46
33 -0.36 -1.25
1728.63
34 -0.29 -1.00
1740.51
35 -0.22 -0.76
1746.42
36 -0.16 -0.53

35
Lag Corr

T

45
LBQ

1748.28
37 -0.09 -0.30
1748.29
38 -0.01 -0.03
1749.70
39 0.08 0.26
1755.14
40 0.15 0.50
1765.20
41 0.20 0.68
1778.55
42 0.23 0.78
1793.44
43 0.24 0.82
1810.08
44 0.25 0.86
1831.09
45 0.28 0.96
1860.92
46 0.34 1.13
1903.36
47 0.40 1.34
1956.81
48 0.45 1.49

Lag Corr

T

LBQ

49 0.48 1.562017.77

Partial Autocorrelation

P artial A utocorrelation Function for y(t)
1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0

5
Lag PAC

T

0.93
-0.46
0.33
-0.14
0.12
-0.13
0.00
-0.06
-0.08
-0.04
-0.04
-0.09

13.08
-6.44
4.71
-1.97
1.72
-1.78
0.01
-0.81
-1.11
-0.55
-0.58
-1.21

1
2
3
4
5
6
7
8
9
10
11
12

15

Lag PAC
13
14
15
16
17
18
19
20
21
22
23
24

-0.05
0.05
-0.07
-0.09
-0.04
-0.08
-0.12
-0.11
-0.15
-0.07
-0.06
-0.15

25

T

Lag

PAC

T

-0.75
0.67
-0.95
-1.30
-0.55
-1.11
-1.71
-1.50
-2.12
-0.93
-0.87
-2.12

25
26
27
28
29
30
31
32
33
34
35
36

-0.03
-0.03
0.00
0.08
-0.01
-0.04
0.15
0.06
-0.00
0.04
0.03
0.04

-0.46
-0.44
0.04
1.13
-0.14
-0.56
2.17
0.88
-0.05
0.49
0.46
0.50

35
Lag PAC
37
38
39
40
41
42
43
44
45
46
47
48

0.08
0.11
0.04
0.04
-0.05
-0.05
-0.14
-0.01
0.07
0.06
-0.01
-0.09

45
T

1.16
1.56
0.51
0.52
-0.74
-0.66
-1.95
-0.10
0.92
0.79
-0.17
-1.24

Lag PAC
49

0.07

T
1.00

‫ﻧﻼﺣﻆ ﻣﻦ ﺷﻜﻞ اﻟﻤﺘﺴﻠﺴﻠﺔ و اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ اﻧﻬﺎ اﺻﺒﺤﺖ ﻣﺴﺘﻘﺮة‬
. d=٢ ‫ﻓﻲ اﻟﻤﺘﻮﺳﻂ اي ان‬
‫ﻣﻦ اﻧﻤﺎط اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ ﻧﺮى اﻧﻬﺎ ﺗﺘﺨﺎﻣﺪ ﻣﻦ اﻟﺘﺨﻠﻒ اﻷول ﻣﻤﺎ‬
‫ وﺳﻮف ﻧﻄﺒﻖ هﺬا اﻟﻨﻤﻮذج ﺑﺎﻷﻣﺮ‬zt ‫ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻷﺻﻠﻴﺔ‬ARIMA(١،٢،١) ‫ﻳﺮﺷﺢ ﻧﻤﻮذج‬
MTB > ARIMA 1 2 1 'z(t)' 'RESI2' 'FITS2';
SUBC>
NoConstant;
SUBC>
Forecast 10 c4 c5 c6;
SUBC> GACF;
SUBC> GPACF;
SUBC> GHistogram;

١٩٩

SUBC>

GNormalplot.

ARIMA Model
ARIMA model for z(t)
Estimates at each iteration
Iteration
SSE
Parameters
0
2462.77
0.100
0.100
1
1345.58
0.250
-0.050
2
1170.63
0.203
-0.200
3
984.83
0.182
-0.350
4
782.47
0.200
-0.500
5
560.15
0.278
-0.650
6
363.93
0.428
-0.765
7
259.20
0.578
-0.814
8
202.76
0.728
-0.842
9
185.51
0.861
-0.859
10
185.36
0.873
-0.860
11
185.36
0.875
-0.860
12
185.36
0.875
-0.860
Relative change in each estimate less than
Final Estimates of Parameters
Type
Coef
StDev
AR
1
0.8749
0.0353
MA
1
-0.8599
0.0357

0.0010

T
24.75
-24.12

Differencing: 2 regular differences
Number of observations: Original series 200, after differencing 198
Residuals:
SS = 183.717 (backforecasts excluded)
MS =
0.937 DF = 196
Modified Box-Pierce (Ljung-Box) Chi-Square statistic
Lag
12
24
36
Chi-Square
3.9(DF=10)
13.0(DF=22)
33.1(DF=34)

48
46.0(DF=46)

Forecasts from period 200
Period
201
202
203
204
205
206
207
208
209
210

Forecast
-22615.2
-22757.2
-22900.4
-23044.5
-23189.5
-23335.2
-23481.5
-23628.4
-23775.8
-23923.7

95 Percent Limits
Lower
Upper
-22617.1
-22613.3
-22764.6
-22749.9
-22917.2
-22883.5
-23075.3
-23013.7
-23238.8
-23140.1
-23407.8
-23262.6
-23582.1
-23380.9
-23761.8
-23495.0
-23946.7
-23605.0
-24136.6
-23710.7

zt = 0.875t −1 z + at − 0.859at −1 , at

N ( 0,0.937 )

Actual

‫اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬

‫ هﻲ‬t ‫وﻣﻘﺪرات اﻟﻤﻌﺎﻟﻢ وإﻧﺤﺮاﻓﺎﺗﻬﺎ اﻟﻤﻌﻴﺎرﻳﺔ وﻗﻴﻤﺔ إﺧﺘﺒﺎر‬

٢٠٠

( )
s.e. (θˆ ) = 0.0357,

φˆ1 = 0.8749, s.e. φˆ1 = 0.0353, t = 24.75
θˆ1 = −0.8599,

1

t = −24.12

σˆ 2 = 0.937, with d . f . = 196

.‫ﻧﻼﺣﻆ ان اﻟﻤﻌﺎﻟﻢ ﻣﻌﻨﻮﻳﺔ‬
:‫اﻵن ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬
ACF of Residuals for z(t)
(with 95% confidence limits for the autocorrelations)
1.0
0.8

Autocorrelation

0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
5

10

15

20

25

30

35

40

45

Lag

PACF of Residuals for z(t)
(with 95% confidence limits for the partial autocorrelations)
1.0

Partial Autocorrelation

0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
-1.0
5

10

15

20

25

30

35

40

45

Lag

‫أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ ﺗﺪل ﻋﻠﻰ أن اﻟﺒﻮاﻗﻲ ﺗﺘﺒﻊ ﺗﻮزﻳﻊ ﺿﺠﺔ‬
:‫ ﻟﻨﻔﺤﺺ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬،‫ﺑﻴﻀﺎء أي ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ‬
‫رﺳﻢ اﻟﻤﺪرج اﻟﺘﻜﺮاري‬

٢٠١

Histogram of the Residuals
(response is z(t))

Frequency

20

10

0
-3

-2

-1

0

1

2

Residual

.‫ﻳﺒﺪو ﻣﺘﻨﺎﻇﺮ ﺑﻌﺾ اﻟﺸﻴﺊ‬
:‫ﻟﻨﻨﻈﺮ إﻟﻰ ﻣﺨﻄﻂ اﻹﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ ﻟﻠﺒﻮاﻗﻲ‬
Normal Probability Plot of the Residuals
(response is z(t))

2

Residual

1

0

-1

-2

-3
-3

-2

-1

0

1

2

3

Normal Score

.‫ﻧﺴﺘﻄﻴﻊ أن ﻧﻘﻮل ان اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ ﺗﻘﺮﻳﺒﺎ‬
.‫ ﻓﺘﺮات ﺗﻨﺒﺆ‬٩٥٪ ‫ ﺗﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻣﻊ‬١٠ ‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟـ‬
Forecast of 20 Future values with 95% limits
-22500

z(t)

-23000

-23500

-24000

0

1

2

3

4

5

6

7

8

9

Time

٢٠٢

10

٢٠٣

:‫اﻟﻤﺮاﺟﻊ‬
١-

Abraham, B. and Ledoter, J. (1983). Statistical
Methods for Forecasting, John Wiley, New York.
٢Anderson, T. W. (1971). The Statistical Analysis of
Time Series, John Wiley, New York.
٣Box, G. E. P. and Jenkins, G. M. (1976). Time Series Analysis

Forecasting and Control, 2nd ed., Holden-Day, San Francisco.
Brillinger, D. R. (1975). Time Series: Data Analysis and
Theory, Holt, Rinehart and Winston, New York.
Fuller, W. A. (1976). Introduction to Statistical Time Series,
John Wiley, New York.
Granger, C. W. J. and Newbold, P. (1977). Forecasting
Economic Time Series, Academic Press, New York.
Hannan, E. J. (1970). Multiple Time Series, John Wiley, New
York.
Harvey, A. C. (1980). Time Series Models, Halsted Press, New
York.
P

٤٥٦٧٨-

P

٩-

Montgomery, D. C., Johnson, L. A. and Gardiner, J. S.
(1990). Forecasting and Time Series Analysis, 2nd ed.,
McGraw-Hill International Edition.
١٠Makridakis, S., Wheelwright, S. C. and McGee, V.
E. (1983). Forecasting Methods and Applications, 2nd ed.,
P

P

P

P

John Wiley, New York.
Shumway, R. H. (1988). Applied Statistical Time Series
١١Analysis, Prentice-Hall, New York.
Wei, W. W. S. (1990). Time Series Analysis Univariate and
١٢Multivariate Methods, Addison Wesley.
١٣Minitab Reference Manual, Release 11 for Windows.
(1998).

T

‫ﻟﻸﺳﻒ اﻟﺸﺪﻳﺪ ﻻﺗﻮﺟﺪ ﺣﺴﺐ ﻋﻠﻤﻲ ﻣﺮاﺟﻊ ﻋﺮﺑﻴﺔ ﺗﻐﻄﻲ آﻞ أو ﺟﺰء ﻣﻦ ﻣﺤﺘﻮى‬
‫اﻟﻤﺎدة اﻟﻤﻐﻄﺎة ﻓﻲ هﺬا اﻟﻜﺘﺎب وأرﺟﻮا ﻣﻦ أي ﻃﺎﻟﺐ أو ﺑﺎﺣﺚ أو ﻣﺪرس ﻳﻌﻠﻢ ﺑﻤﺜﻞ‬
:‫هﺬا اﻟﻤﺮﺟﻊ او اﻟﻜﺘﺎب أن ﻳﺮﺳﻞ ﻟﻲ ﻣﻼﺣﻈﺔ ﻋﻠﻰ اﻟﺒﺮﻳﺪ اﻹﻟﻜﺘﺮوﻧﻲ‬
abarry@ksu.edu.sa ‫ أو‬abarry@abarry.net
.‫وﺷﻜﺮا‬
T

T

T

٢٠٤