Journal of Mathematics, Год журнала: 2024, Номер 2024(1)
Опубликована: Янв. 1, 2024
When combining numerous individual forecasting models linearly, it is common to construct a measure of model validity based on AD() or APE(). Specifically, we generally by using the mean and standard deviation AD APE. The resulting from this approach will serve as basis for assigning weights, then weights be used combine multiple into linear (LCFM).However, drawback APE alone that they do not consider varying importance between recent errors long‐term errors. To address limitation, article reconstructs new type (New Model Validity). sliding average fitting accuracy are two components New Validity, enhances incorporating smoothing coefficient. Thus when applying Validity LCFM, assigned higher weight if its more accurate in period. Through proof, shown long appropriate obtained, LCFM remains superior each under Validity. Therefore, feasible attempt improve performance
Язык: Английский