factory.gif (3606 bytes)

QM 3523 -- Operations Management
Department of Economics & Decision Sciences
torch1.gif (1833 bytes)asu.gif (4934 bytes)
College of Business


[Home]  [Contents]  [Syllabus]  [Homework]  [Handouts]  [Grades]  [Excel]  [Old Exams]  [Miscellaneous]


Time Series Analysis Example

     Refer to the historical demand for Product 4 in "Data Analysis Exercise #2," available from the Handouts page of the course web site.  Four different forecasting techniques were used to analyze the data.  For each, the optimal model parameters (i.e., b0 and b1 for trend, SI for seasonal index, and a for exponential smoothing) were determined, and the resulting models were used to predict past demand, as well as demand for time period 13.  The table below summarizes the results of the analyses.  Note that a number of a values were used with the exponential smoothing approach, and the value of 0.95 resulted in the lowest MAD.  

     Clearly, simple exponential smoothing is not an appropriate forecasting approach for this product.  (Exponential smoothing should be used only when no trend or seasonality are exhibited, and this example demonstrates what would happen if this advice were ignored.)  It should also be clear that the best approach to use for forecasting future demand for Product 4 is the pure trend approach, since it resulted in the lowest MAD when used to predict the past.  Hence, our best forecast of demand for first quarter next year (i.e., Period 13) is 41,800 units, and we expect this to be wrong by somewhere around 1900 units (the trend model's MAD).

 
Period Demand Trend Seasonal Combined Exponential
t At Ft |et| Ft |et| Ft |et| Ft |et|
1
2
3
4
5
6
7
8
9
10
11
12
15
17
24
28
24
29
31
30
31
37
37
40
17.4
19.4
21.5
23.5
25.5
27.6
29.6
31.6
33.7
35.7
37.7
39.8
2.41
2.44
2.53
4.50
1.54
1.43
1.40
1.63
2.66
1.31
0.72
0.24
23.0
27.1
31.0
33.3
23.0
27.1
31.0
33.3
23.0
27.1
31.0
33.3
  7.98
10.13
  6.98
  5.25
  1.02
  1.87
  0.02
  3.25
  8.02
  9.87
  6.02
  6.75
15.4
19.8
24.5
28.3
20.9
26.4
32.0
36.3
26.5
32.9
39.5
44.4
0.37
2.79
0.47
0.28
3.07
2.64
0.98
6.33
4.50
4.06
2.48
4.39

15.0
15.2
16.1
17.3
17.9
19.1
20.3
21.2
22.2
23.7
25.0

  2.00
  8.80
11.92
  6.73
11.06
11.95
  9.75
  9.78
14.80
13.32
14.99
13 n/a 41.8 n/a 23.0 n/a 32.1 n/a 26.5 n/a
Best method:

  Pure Trend  

b0 = 15.38
b1 = 2.03
R2 = 91%
Syx = 2.39
SI1 = 0.804
SI2 = 0.949
SI3 = 1.084
SI4 = 1.163
b0 = 17.38
b1 = 1.73
R2 = 77%
Syx = 3.62
Alpha = 0.95
MAD = 1.90 MAD = 5.60 MAD = 2.69 MAD = 3.23


Back to Top


(This page updated on 01/20/00 )

Please submit any comments, corrections, etc. about this document to John Seydel

QM 3523 Home Page