An Evaluation of Rules for Selecting an Extrapolation Model on Yearly Sales Forecasts

Six rules are used to select among seven exponential smoothing models (to pick the model that yields the lowest error on a fit to historical data). Comparisons of accuracy are made among forecasts generated by the selection rules and the individual smoothing models. The results indicate that the use of rules leads to increases in accuracy. A combination of selection rules led to the most accurate forecasts. The rules also led to lower variances in the accuracy of forecasts, that is, they help to avoid large errors.