E-commerce customer churn prediction based on improved SMOTE and AdaBoost

E-commerce customer churn rate is high and the customer churn dataset is seriously imbalanced. In order to improve the prediction accuracy of churn customers as well as strengthen to identify non-churn customers, this paper presents e-commerce customer churn prediction model based on improved SMOTE and AdaBoost. First, processing the churn data with improved SMOTE, which combines oversampling and undersampling methods to address the imbalance problem and subsequently integrates AdaBoost algorithm to predict. Finally, the empirical study on B2C E-commerce platform proves that this model has better efficiency and accuracy compared with the mature customer churn prediction algorithms.