Professional Documents
Culture Documents
Amit Tyagi
Harmanjit Singh
Mohini Jain
Nandini Chandrasekhar
Rahul Chakraborty
OBJECTIVE
Objective
Data
DATA
Background
Quick Look
Issues
Preparation
ANALYSIS
Analysis
Naive
Exponential Moving Average
Time Series Forecasting
Stakeholders
STAKEHOLDERS
Time
TIME Period:
PERIOD Data available for the period Jan 2009 – May 2011
Forecast
FORECASTPeri June 2011 – Aug 2011
Based on monthly prices of cement and oil for the years 2009 and 2010
Average Price of Cement in Major Consumption (Centers Per Bag of 50 kg (In Rs.241)
Monthly Average Price of Indian Basket of Crude Oil (in $/barrel converted to Rs3471)
Moving Average
Missing data imputation in cement prices using the method of Moving
Average and straight line method
Oil prices converted to Indian Rupees using historic conversion rate
corresponding to each data point sourced from www.oanda.com
Cement
• Level: Averages around 237.5
• Trend: No clear Trend present in the series
• Seasonality: No significant pattern
• Stationary series
• Peak around the month of April every year
Oil
• Level: for oil prices is around 3500
•Trend: Clear increasing trend
• Seasonality : No significant pattern
• Non Stationary Series
• 2010 values follow a very smooth pattern
Cement data displays monthly seasonality
Naïve 12-month ahead forecast
Forecast values and Error Measurements attached below
290.00
Forecasted
280.00 Time Cement Prices
270.00 June. 2011 240.00
260.00
July. 2011 233.00
Aug. 2011 225.00
250.00
240.00
270
260 Validation method systematically under
250 Period
240
230
predicts the cement demand
220
210
200
MAPE -0.51%
MAE 19.24
Time
RMSE 10.78
Actual Forecast
260
250 •The RMSE error is lower for the method
240
Test Data
230
than previous methods.
220 • However the method systematically
210 under predicts in validation set
200
• Adjusted R 2 for the model is .3345
July. 2010
Mar. 2009
July. 2009
Sept. 2009
May. 2010
Jan. 2010
Nov. 2010
May. 2009
Nov. 2009
Mar. 2010
Sept. 2010
Jan. 2009
290
Validation Data
270
250
230
210
Jan. 2011 Feb. 2011 Mar. 2011 Apr. 2011 May. 2011
Independent Variables : time & Dummy variables for each Month( Dec. as Base Month)
260
Test
250 •The RMSE error is lower for the method
240
Data than previous methods.
230
220 •Method has similar errors in validation
210 and test sets
200 •Adjusted R 2 for the model is .5364
300
280
Validation 260
Data
240
220
Jan. 2011 Feb. 2011 Mar. 2011 Apr. 2011 May.
2011
Independent Variables : Change in Crude oil & Dummy variables for each Month( Dec. as
Base Month)
Test 260
•The RMSE error is lower for the method
Data 240 than previous methods.
220 •Method systematically under predicts in
200
validation/Test Set
•Adjusted R 2 for the model is .5964
290
280
270
Validation 260
Data
250
240
230
220
Jan. 2011 Feb. 2011 Mar. 2011 Apr. 2011 May. 2011
Based on Model 3 the forecast for June 2011 is:
Point Estimate: 244.67
95% confidence interval is (230.50,258.9)
-0.5
-1
1 2 3 4 5 6
Number of Lags
INSIGHTS RECOMMENDATIONS