Can Online User Behavior Improve the Performance of Sales Prediction in E-commerce?

Abstract

How to forecast product sales effectively and efficiency in E-commerce is a significant task for E-commerce producers to manage product inventory and design marketing strategies. However, under the uncertainty of product demand, sales prediction is a complex task. This paper presents a novel data mining framework for sales prediction based on online user behavior data. Under the framework, the relationship of sales data and online user behavior data is well modelled, and the optimal lag of online user behavior data for sales prediction is also identified. In terms of evaluation criteria, a number of books are used for sales prediction. The empirical results show the efficiency and effectiveness of the proposed framework and also revealed that among different categories of books, the forecasting performance of some categories including Finance and Exam heavily relies on online user behavior information. So, it indicates that the proposed framework can be used as a potential alternative to analyze the sales trend, and help managers in Ecommerce companies for inventory optimization and customer relationship management.