Churn Prediction With Explainability for the Customers of Telecom Industry

Customer churn prediction is an essential task in any market segment that deals with customers. The Telecom industry is one such market sector where customer management plays a crucial role. From existing research, it is clear that the cost of retaining the current customers is much lower than the cost of gaining new customers. Hence, the prior detection of the churn customers becomes very important. This paper presents a churn prediction methodology for the telecom industry that aims at better identifying the customers who might churn. Seven machine learning models are trained using three publicly available datasets. The experimental results of this study reveal great performances compared to existing works in terms of recall, F1-score, and AUC-score. Moreover, new feature crosses are explored in this research which are found to be impactful in predicting customer churn. Finally, the best models are explained using SHapley Additive exPlanations (SHAP), an explainable AI technique.