Article ID: | iaor20003241 |
Country: | United States |
Volume: | 45 |
Issue: | 8 |
Start Page Number: | 1169 |
End Page Number: | 1176 |
Publication Date: | Aug 1999 |
Journal: | Management Science |
Authors: | Gardner Everette S. |
This paper evaluates the ex ante performance of rule-based time series forecasting systems proposed in earlier research. The author shows that comparable performance can be obtained with a simpler alternative, a damped-trend version of exponential smoothing fitted to minimize the Mean-Absolute-Deviation (MAD) criterion. The results suggest that the performance of rule-based systems would be improved through this alternative and that time series forecasters should consider MAD fits in model development.