Research of CLV Forecasting Model Based on E-Commerce Website

To predict the customers´ transaction behavior is essential to the basic work of client management in BtoC website. The Pareto/NBD model was generally accepted in this field. The BG/NBD model is an improved model of the Pareto/NBD. The paper makes use of actual transaction data at a site in Liaoning province to verify the applicability of the BG/NBD in this field and redefines some variables more suitable for their practical application. By comparison with the actual data through diagram, the results show as follows. The prediction outcome of the BG/NBD model is very close to the actual data in the high trading frequency, despite a certain deviation compared with the actual in low frequency; The BG/NBD model can capture the hidden information in the sales data and be effectively used in China´s online marketing to forecast customer buying behavior; in addition, this article summarizes the method to set warning point for customer loss rate.