For each retail series, Table 1 gives the forecasting accuracy
measures for in-sample data of the ETS model and ARIMA model
selected according to the procedure described earlier. The forecast
error measures presented in Table 1 are defined in Section 3.1.
From Table 1 it can be observed that with the exception to Shoes
series ARIMA models forecast better than ETS models in the
training sample judged by the three most common performance measures: RMSE,MAE and MAPE (with the single exception to MAPE ของ Sandals series); although it should be emphasized that these results should not be used for forecast evaluation.