Further Extensions of Adaptive Extended Exponential Smoothing and Comparison with the M-Competition

Abstract

This study presents and evaluates three new approaches to nonintervention, extrapolative (time series) forecasting. This is an extension of the adaptive extended exponential smoothing methodology (AEES) that allows the model additional smoothing constant adaptability to improve forecasting accuracy. The performance of the basic AEES method and two enhancements are first compared to five other time series techniques on a limited, validation data set and then compared to the 24 methods used in the M-Competition. Comparisons are made across all 111 M-Competition data sets and across the yearly, quarterly, and monthly components of the 1]] data sets. When empirically tested across the III M-Competition data series, the heuristic AEES approach generally provided improved or comparable accuracy. This result was repeated with the yearly data series. Results for the quarterly and monthly data series were mixed. Discussion of these results within the marketing context of sales forecasting is provided.