Professional Documents
Culture Documents
IE3265 Forecasting
IE3265 Forecasting
Chapter 2
IE 3265
R. Lindeke, Ph. D.
Introduction to Forecasting
What is forecasting?
Characteristics of Forecasts
It should be timely
It should be as accurate as possible
It should be reliable
It should be in meaningful units
It should be presented in writing
The method should be easy to use
and understand in most cases.
Customer Surveys
Jury of Executive Opinion
The Delphi Method
Causal Models
Let Y be the quantity to be forecasted and (X1,
X2, . . . , Xn) are n variables that have predictive
power for Y. A causal model is Y = f (X1, X2, . . . ,
Xn).
A typical relationship is a linear one:
Y = a0 + a1X1 + . . . + an Xn
Notation Conventions
Evaluation of Forecasts
MAD = (1/n) | e i |
MSE = (1/n) ei 2
Evaluation of Forecasts
Moving Averages
Demand
Month
Demand
January
89
July
223
February
57
August
286
March
144
September
212
April
221
October
275
May
177
November
188
June
280
December
312
Summary of Moving
Averages
Easily understood
Easily computed
Provides stable forecasts
1.
2.
{Or : Ft + 1 = Ft - Ft - Dt) }
Weights in Exponential
Smoothing:
Jan 23.3
Feb 72.3
Mar 30.3
Apr 15.5
And the January Forecast was: 25
Using = .15
Forecast for Feb: Djan + (1- )Fjan = .15*23.3 +
(.85)*25 = 24.745
Forecast for Mar: Dfeb + (1- )Ffeb = .15*72.3 +
(.85)*24.745 = 31.88
Apr: Dmar + (1- )Fmar = .15*30.3 + .85*31.88 = 31.64
May: Dapr + (1- )Fapr = .15*15.5 + .85*31.64 = 29.22
Comparison of MA and ES
Similarities
Comparison of MA and ES
Differences
Model: Dt = a + bt + t.
If t is scaled to 1, 2, 3, . . . , -- it becomes a
number i -- then the least squares estimates
for a and b can be computed as follows: (n is the
number of observation we have)
S xx
S xy n iDi
n * (n 1)
i 1
n 2 n 1 2
n 2 (n 1) (2n 1
Di
2 *
i 1
n 1
a D b
S xy
S xx
# Visitors
133
183
285
Month
Apr
May
Jun
# Visitors
640
1875
2550
Sxy=6*(1*133+2*183+3*285+4*640+5*1875+6*2550)
(6*7/2)(133+183+285+640+1875+2550)] = 52557
Sxx = [(36*7*13)/6]-[(36*49)/4)]= 105
b = (52557/105)= 500.54
a = 944.33 500.54*(6+1)/2 = -807.4
Deseasonalizing a Series
Lets do one:
Season
Cycle
Deman
d
Season
Cycle
Deman
d
Q1
2001
205
Q1
2002
225
Q2
2001
225
Q2
2002
248
Q3
2001
185
Q3
2002
203
Q4
2001
285
Q4
2002
310
215
236.5
194
298
Lets do one:
Q1: 215/235.75 =
Q2: 236.5/235.75 =
Q3: 194/235.75 =
Q4: 297.5/235.75 =
Sum:
0.912
1.003
0.823
1.262
4.000
Deseasonalize the
Demand
Demand
Avg.
Factor
Demand
X2
string (x)
(y)
Q1 01
205
215
.912
224.78
224.78
Q2 01
225
236.5
1.003
224.29
448.57
Q3 01
185
194
.823
224.81
674.43
Q4 01
285
298
1.262
225.84
903.36
16
Q1 02
225
215
.912
246.72
1233.6
25
Q2 02
248
236.5
1.003
247.21
1483.26
36
Q3 02
203
194
.823
246.64
1726.48
49
Q4 02
310
298
1.262
245.66
1965.28
64
SUM:
8659.76 204
Using Deseasonalized
Tableau
xy n x y
x n x
2
a=
y b* x
Finally:
Q1, 03
a+bx
217.24+4.113*9
254.26 *.912
231.9 23
2
Q2, 03
a+bx
217.24+4.113*1
0
258.37*1.003
259.2 25
9
Q3, 03
a+bx
217.24+4.113*1
1
262.48*.823
216.0 21
6
Q4, 03
a+bx
217.24+4.113*1
2
266.6*1.262
336.4 33
6
The Trend:
The Seasonal
Factors:
Gt St St 1 1 Gt 1
ct
Dt
1 ct n
St
Beginning Winters
Method:
Typically we set: = 2 = 2
Deriving initial estimates takes at
least two complete cycles of data
V1 1 D j
Compute sample
n j 2 n1
means for each cycle of average demand for 2 cycles 'ago'
data (V1 and V2)
0
V2 1
n1
n j
Dj
G0
V2 V1
Signal:
Seasonal Factor ct
Estimates:
Dt
n 1
Vi 2 j G0
Averaging ci:
c2 n1 c n1
c n1
2
Normalizing:
cj
cj
c i
i n 1
denominator is sum of factors
0
Obs.
Demand
72
107
55
88
Period
1st Q
(02)
2nd Q
(02)
3rd Q
(02)
4th Q
(02)
Obs.
Demand
83
121
63
100
4 72 107 55 88 80.5
V 1 83 121 63 100 91.75
4
V1 1
2
V2 V1
(91.75 80.5 2.813
4
4
G0
4 1
S0 V2 G0
95.97
ct
Dt
n 1
j G0
Vi
83
91.75 5 1 2.813
2
.944 .948
avg :
2 .946
.948
(1st
Same approach:
Seasons 2, 3
&4
Normalizing:
n
cj
cj
n 1
c i
i 0
denominator is sum of factors
Here:
Making a Forecast:
Ft ,t St Gt c t n
F0,1stQ S0 1 G0 c1stQ
S1
.2 112
.944
1 S0 G0
c1stQ
G1 S1 S0 1 G0
c1stQ
D1
1 c1stQ
S1
.1 112
.9 .944 .959
102.76
Lets Try 1:
Practical Considerations