CUSTOMER CHURN PREDICTION AND CATEGORIZATION A MACHINE LEARNING APPROACH TO ANALYSE CUSTOMER BEHAVIOR AND DECISION MAKING IN THE TELECOMMUNICATIONS INDUSTRY

Customer churn remains a critical challenge for subscription- based businesses, particularly in the telecommunications indus- try, where retaining customers is significantly more cost-effective than acquiring new ones. This study leverages machine learn- ing to develop a robust churn prediction framework and identify key behavioral drivers of churn. Using the Telco customer churn dataset, we employ an ensemble Voting Classifier composed of Logistic Regression, Random Forest, XGBoost, and CatBoost models. The ensemble achieves a high predictive accuracy, with an AUC of 0.98, effectively distinguishing between churned and retained customers. Beyond prediction, the study introduces a structured catego- rization of churn reasons into four primary classes-Attitude and Expertise, Service and Product Issues, Competitor and Price, and Other Reasons. A multi-class classification model using XGBoost achieves an accuracy of 0.63, outperforming random guess baselines. The categorization reveals that factors such as short-term contracts, lack of technical support, and high monthly charges are significant contributors to churn. Conversely, long- term contracts and automated payment methods demonstrate strong retention effects. The findings provide actionable insights into customer behav- ior and decision-making, emphasizing the importance of improv- ing technical support, addressing cost concerns, and creating in long-term commitments. By combining predictive accuracy with interpretability, this study enables targeted retention strategies to minimize churn and enhance customer lifetime value.