A Robust Machine Learning System for Classification and Prediction of Customer Churn in Telecom Sector

Customer desertion detection is one of the most critical research topics that the telecommunications industry must address in order to retain existing customers. Churn is the term utilised to describe a loss of consumers as a result of the existing offers of competitors or, in some cases, network issues. Customers may prefer to terminate their subscriptions to services in these circumstances. Customer churn prediction is crucial for retaining long-term customers and ensuring profitability. The main goal of this system is to analyse various ML algorithms that are required to build models that can anticipate customer attrition and, once retention tactics and plans have been implemented, to identify the reasons behind churn. This study focuses on developing an effective churn prediction model using the Telco customer churn dataset by Kaggle, encompassing 7043 customer records. Through comprehensive data preprocessing, feature selection, and employing ML models—ETC, Decision Tree, and Stacking Classifier—the study evaluates and compares their performance in predicting customer churn according to F1-score, recall, accuracy, and precision. The outcomes demonstrate that a Stacking Classifier achieves the highest accuracy of 91%, highlighting its potential for practical implementation in the telecom sector to reduce churn rates and enhance customer retention strategies.