USING PREVIOUS OBSERVATIONS IN A DECISION RULE FOR SELECTING EITHER CROSTON’S METHOD OR SINGLE EXPONENTIAL SMOOTHING

摘要

Knowing when to use single exponential smoothing versus Croston’s Method when demand patterns change has not been adequately addressed. The literature recommends utilizing a classification scheme to determine the proper forecasting methodology based on predetermined cut off points quantified by the number of periods with no demand. The proposed Hybrid Croston approach outperforms the traditional method and SES in limited simulations and conditions and parameters when the methodology excels are determined. Two different demand rates to represent periods of intermittent and regular demand are utilized. The forecast switches between Croston’s procedure and SES based on a proposed rule.