Operations Management - Evaluation of Each Forecasting Methods
Autor: Cordel Driscoll • March 21, 2016 • Case Study • 791 Words (4 Pages) • 964 Views
All calculations were computed in the attached excel file.
Question 1
- Data Plotted
[pic 1]
Evaluation of each forecasting methods
Naïve Method: This method would not be perform very well in forecasting future sales because the sales fluctuate from month to month resulting first in an upward trend then a slight downward trend then resuming an upward trend at the end of the year. There is also the concern of seasonality in this data and the naïve method does not perform well with this factors.
Moving Average: This forecasting method takes an average of previous periods into consideration as opposed to just the single previous period. By increasing the size of n, which is the number of periods averaged it will smooth out fluctuations but is makes the method less sensitive to real changes in the data. Even with the additional periods used to forecast it is still not an adequate method to use and will not perform well with this data because it does not pick up trend very well and extensive records of past data is require where as we only have 12 months of data.
Simple Exponential Smoothing: This method would perform well with this data because little record keeping of past data is required. The most recent data is weighted heavier, and this method can adjust for stable vs non-stable time series data.
Double Exponential Smoothing: This method will perform the best because the data has trend and the double exponential smoothing method is modified to capture this trend component. As seen in class simple exponential smoothing tends to under forecast values when trend is present as opposed to this method that captures that gap.
b)
Naïve | |||||
t | At | Naïve | ERROR | ABS ERR | SQ ERR |
Jan | 126 |
|
|
|
|
Feb | 137 | 126 | 11 | 11 | 121 |
Mar | 142 | 137 | 5 | 5 | 25 |
Apr | 150 | 142 | 8 | 8 | 64 |
May | 153 | 150 | 3 | 3 | 9 |
Jun | 154 | 153 | 1 | 1 | 1 |
Jul | 148 | 154 | -6 | 6 | 36 |
Aug | 145 | 148 | -3 | 3 | 9 |
Sep | 147 | 145 | 2 | 2 | 4 |
Oct | 151 | 147 | 4 | 4 | 16 |
Nov | 159 | 151 | 8 | 8 | 64 |
Dec | 166 | 159 | 7 | 7 | 49 |
MAD= | 5.27272727 | ||||
MSE= | 36.1818182 | ||||
Moving Average | n=4 | ||||
t | At | M.A | ERROR | ABS ERR | SQ ERR |
Jan | 126 |
|
|
|
|
Feb | 137 |
|
|
|
|
Mar | 142 |
|
|
|
|
Apr | 150 |
|
|
|
|
May | 153 | 138.75 | 14.25 | 14.25 | 203.0625 |
Jun | 154 | 145.5 | 8.5 | 8.5 | 72.25 |
Jul | 148 | 149.75 | -1.75 | 1.75 | 3.0625 |
Aug | 145 | 151.25 | -6.25 | 6.25 | 39.0625 |
Sep | 147 | 150 | -3 | 3 | 9 |
Oct | 151 | 148.5 | 2.5 | 2.5 | 6.25 |
Nov | 159 | 147.75 | 11.25 | 11.25 | 126.5625 |
Dec | 166 | 150.5 | 15.5 | 15.5 | 240.25 |
MAD= | 7.875 | ||||
MSE= | 87.4375 |
Out of these two methods we would recommend to use the Naïve method because both the MAD and MSE values are lower than the Moving average values resulting in more accurate forecasting over this data.
c)
Exponential Smoothing | a=0.1 | ||||
t | At | EXP SM | ERROR | ABS ERR | SQ ERR |
Jan | 126 | 126 |
|
|
|
Feb | 137 | 126 | 11 | 11 | 121 |
Mar | 142 | 127.10 | 14.90 | 14.90 | 222.01 |
Apr | 150 | 128.59 | 21.41 | 21.41 | 458.39 |
May | 153 | 130.73 | 22.27 | 22.27 | 495.91 |
Jun | 154 | 132.96 | 21.04 | 21.04 | 442.77 |
Jul | 148 | 135.06 | 12.94 | 12.94 | 167.39 |
Aug | 145 | 136.36 | 8.64 | 8.64 | 74.72 |
Sep | 147 | 137.22 | 9.78 | 9.78 | 95.64 |
Oct | 151 | 138.20 | 12.80 | 12.80 | 163.88 |
Nov | 159 | 139.48 | 19.52 | 19.52 | 381.09 |
Dec | 166 | 141.43 | 24.57 | 24.57 | 603.66 |
MAD= | 16.2614 | ||||
MSE= | 293.3144 | ||||
Exponential Smoothing | a= | 0.2 | |||
t | At | EXP SM | ERROR | ABS ERR | SQ ERR |
Jan | 126 | 126 |
|
|
|
Feb | 137 | 126 | 11 | 11 | 121 |
Mar | 142 | 128.20 | 13.80 | 13.80 | 190.44 |
Apr | 150 | 130.96 | 19.04 | 19.04 | 362.52 |
May | 153 | 134.77 | 18.23 | 18.23 | 332.41 |
Jun | 154 | 138.41 | 15.59 | 15.59 | 242.91 |
Jul | 148 | 141.53 | 6.47 | 6.47 | 41.84 |
Aug | 145 | 142.83 | 2.17 | 2.17 | 4.73 |
Sep | 147 | 143.26 | 3.74 | 3.74 | 13.99 |
Oct | 151 | 144.01 | 6.99 | 6.99 | 48.89 |
Nov | 159 | 145.41 | 13.59 | 13.59 | 184.78 |
Dec | 166 | 148.13 | 17.87 | 17.87 | 319.51 |
MAD= | 11.68189 | ||||
MSE= | 169.3648 |
Exponential Smoothing | a= | 0.5 | |||
t | At | EXP SM | ERROR | ABS ERR | SQ ERR |
Jan | 126 | 126 |
|
|
|
Feb | 137 | 126 | 11 | 11 | 121 |
Mar | 142 | 131.50 | 10.50 | 10.50 | 110.25 |
Apr | 150 | 136.75 | 13.25 | 13.25 | 175.56 |
May | 153 | 143.38 | 9.63 | 9.63 | 92.64 |
Jun | 154 | 148.19 | 5.81 | 5.81 | 33.79 |
Jul | 148 | 151.09 | -3.09 | 3.09 | 9.57 |
Aug | 145 | 149.55 | -4.55 | 4.55 | 20.67 |
Sep | 147 | 147.27 | -0.27 | 0.27 | 0.07 |
Oct | 151 | 147.14 | 3.86 | 3.86 | 14.92 |
Nov | 159 | 149.07 | 9.93 | 9.93 | 98.64 |
Dec | 166 | 154.03 | 11.97 | 11.97 | 143.18 |
MAD= | 7.623846 | ||||
MSE= | 74.57288 |
0.1 | 0.2 | 0.5 | |
MAD | 16.2614 | 11.68189 | 7.623846 |
MSE | 293.3144 | 169.3648 | 74.57288 |
By comparing a=0.1, 0.2 and 0.5 we would recommend to use a=0.5 because it provides us with the smallest MAD and MSE resulting in the smaller error when forecasting. This result is consistent with our earlier analysis that the time series is not stable.
...