Realization of a demand forecasting algorithm for retail industry

This paper proposes the realization of a demand forecasting algorithm for retail industry. It categorizes the merchandises into four cases according to the availability of the historical demand, and suggests processes to deal with each individual case. This paper adopts Holt-Winters' model for the forecasting algorithm, and provides method to initialize the parameters. Due to the dependency of Holt-Winters' method on the reliability of historical demand collection, the accuracy of the proposed algorithm is limited to a certain degree. However, this applicable process adopts smoothing methods to minimize the influence by unreliable historical demand and could provide reference for the relevant research.