Abstract:

Sales forecasting is very important for decision-making and strategy management of firms. However, sales forecasting is a challenging problem due to the fact that sales are influenced by many internal and external factors. The traditional forecasting models usually use the historical sales and product attributes to predict the future sales, rarely utilizing the sales information of other related products. This paper applied market basket analysis (MBA) to explore the correlation between different categories. Then we constructed the network of the categories and utilized the sales of the associated categories to enhance the prediction accuracy of the focal category. In order to solve the endogeneity problem during the prediction process, we adopted vector auto regression (VAR) to model the forecasting problem. In addition, the influences of the holiday and weekend were incorporated in the model. Our forecast model was applied to the sales data from a supermarket in China. The results demonstrated that the proposed method achieved higher forecasting accuracy than traditional methods. Finally, according to the results, some managerial suggestions were proposed for the inventory management and marketing decision.

Key words: Sales forecasting, Market basket analysis, Social network analysis, Vector auto regression