Wrapper Feature Selection Optimized SVM Model for Demand Forecasting

An accurate demand forecasting model has academic and practical significance to supply chain management for China´s retail industry. In this paper, a novel demand forecasting model named WFSSVM (Wrapper Feature Selection optimized SVM) is proposed. Genetic algorithm based wrapper feature selection method is firstly employed to analyze the sales data of a kind product (including various kinds of brand). Then, the selection result is applied to build Support Vector Machine (SVM) regression model. Different other approaches such as Winter Model, Radius Basis Function Neural Network (RBFNN) and SVM without feature selection are also used for comparison and evaluation. The final experiment result proves the efficiency of the model.