Fuzzy Forecast Combining for Apparel Demand Forecasting

Abstract

In this chapter, we present a novel approach for apparel demand forecasting that constitutes a main ingredient for a decision support system we designed. Our contribution is twofold. First, we develop a method that generates forecasts based on the inherent seasonal demand pattern at product category level. This pattern is identified by estimating lost sales and the effects of special events and pricing on demand. The method also allows easy integration of product managers’ qualitative information on factors that may affect demand. Second, we develop a fuzzy forecast combiner. The combiner calculates the final forecast using a weighted average of forecasts generated by independent methods. Combination weights are adaptive in the sense that the weights of the better-performing methods are increased over time. Forecast combination operations employ fuzzy logic. We illustrate our approach with a simulation study that uses data from a Turkish apparel firm.