With the rapid development of information technology, most enterprises have built e-commerce platform, which promotes the revolution of operation mode. The focus of competition gradually becomes the customers rather than the products under the increasingly fierce market competition of the E-commerce model. Because of the non-contractual relationship between the customers and the e-commerce platform, maintaining the stable customer relationship becomes the necessary condition for the e-commerce enterprises to get profit. So predicting the customer churn accurately plays an important role in the development of e-commerce enterprises. In this paper, the BG / NBD model is used to analyze the historical transaction records of an e-commerce platform in order to analyze and predict the purchase behavior of the existing customers, and identify the pre-losing customers, which helps the enterprises to implement the more effective strategies of CRM and restore the pre-loss customers timely.