The HPFESMSPEC Procedure

(View the complete code for this example.)

This example illustrates how to define model specifications that automatically choose the best exponential smoothing model by using MAPE as the model selection criterion.

The following statements fit two forecast models (simple and log simple exponential smoothing) to the time series. The forecast model that results in the lowest MAPE is used to forecast the time series.

proc hpfesmspec repository=mymodels
                name=best_simple;
   esm method=simple transform=auto criterion=mape;
run;

The following statements fit two forecast models (seasonal and log seasonal exponential smoothing) to the time series. The forecast model that results in the lowest MAPE is used to forecast the time series.

proc hpfesmspec repository=mymodels
                name=best_seasonal;
   esm method=addseasonal transform=auto criterion=mape;
run;

The following statements fit 14 forecasting models (best and log best exponential smoothing) to the time series. The forecast model that results in the lowest MAPE is used to forecast the time series.

proc hpfesmspec repository=mymodels
                name=best;
   esm method=best transform=auto criterion=mape;
run;