IMPLEMENTATION AND EVALUATION OF CUSTOMER CHURN PREDICTION MODELS FOR E-BUSINESS SERVICES USING XGBOOST

Abstract

The substantial growth of e-business streaming platforms has led to increased competition and difficulty with subscriber retention and attrition. Particularly concerning is the inability of many platforms to predict subscriber loss. The present study is intended to build and empirically assess a predictive model to detect potential subscriber loss to allow e-business streaming platforms to undertake more focused subscriber retention actions. The proposed model is based on the Extreme Gradient Boosting (XGBoost) method. Predictive e-business analytics based on customer engagement, subscription type, and subscriber demographics will consist the predictors. The model will be trained, tested and fine-tuned on the e-business predictive analytics data and the classification and assessment of model performance will be based on the accuracy, precision, recall, F1, and AUC metrics. The performance of the model based on the assigned e-business predictive analytics data shows a marked improvement over existing predictive methods and techniques. The new model will serve as a decision-making tool aimed at churn reduction and improvement of customer lifetime profitability for e-business streaming services. These results underscore the value of machine learning approaches for predictive analytics in subscriber retention and business profitability.