A Hybrid Prediction Model for E-Commerce Customer Churn Based on Logistic Regression and Extreme Gradient Boosting Algorithm

Received: 17 April 2019 Accepted: 23 July 2019 Customer churn is an important problem in the field of e-commerce. Based on the real data of an e-commerce platform, this paper establishes a hybrid prediction model for customer churn based on logistic regress and extreme gradient boosting (XGBoost) algorithm. More than 20 key indices were selected through data mining of the real data, covering such dimensions as order information, customer profile, and aftersales situation. With these indices, the hybrid model was applied to predict the churn state of each customer in the sample data. The results show that our model achieved a greater-than-85% accuracy in the forecast of customer churn. The research findings provide an important guide for e-commerce enterprises to improve customer adhesiveness.