Transfer Ensemble Model for Customer Churn Prediction with Imbalanced Class Distribution

Customer churn prediction is an important issue in customer relationship management. The class distribution of customer data is often imbalanced, which may affect the performance of churn prediction model greatly. This paper combines transfer learning and multiple classifiers ensemble, and proposes a transfer ensemble model for imbalanced data (TEMID). This method focuses on using transfer learning and sampling to enlarge the available training set and balance it respectively. What´s more, it also uses multiple classifiers ensemble method to implement the classification. The performance of TEMID and some existing transfer learning algorithms are compared in two class imbalanced datasets. The results show that the TEMID methods can actually improve the performance of the customer churn prediction.