Development of a Customer Churn Prediction Model Using Machine Learning Techniques in the Telecommunications Industry

Abstract

Customer churn remains a major challenge in the telecommunications industry due to increasing market competition and customer mobility. This study developed and evaluated machine learning models for predicting customer churn using the Telco Customer Churn dataset containing 7,043 customer records. The study applied data preprocessing techniques including missing value handling, categorical encoding, and feature scaling before implementing Logistic Regression and Random Forest classification models. Model performance was evaluated using Accuracy, Precision, Recall, F1-Score, and ROC-AUC metrics. Experimental results showed that the Random Forest classifier achieved superior predictive performance, with an accuracy of 80%, a recall of 57%, an F1-score of 0.62, and an ROC-AUC of 0.85. Feature importance analysis revealed that contract type, tenure, monthly charges, and total charges were the most significant predictors of customer churn. The findings demonstrate the effectiveness of machine learning techniques in supporting proactive customer retention strategies and data-driven decision-making in the telecommunications sector.