Prediction of Banking Customer Churn Based on XGBoost with Feature Fusion

With the increasing competition in the banking industry, accurate prediction of banking customer churn has become an important way in managing customer relationships. To explore efficacy features, enhance the generalization performance of customer churn prediction, this study proposed a XGBoost model with feature fusion for banking customer churn prediction. At first, a feature fusion model based on improved RFM and Affinity Propagation clustering was proposed to extract features representing the long-term and dynamic behavior of customers. By integrating different types of features, a XGBoost model was proposed to predict customer churn. Experimental results demonstrate the superior performance of the proposed model in comparison to other benchmark models.