Customer Churn Prediction Using Machine Learning Models: A Comparative Study Using R Software
TL;DR
This study compared explainable machine learning models for predicting customer churn using the IBM Telco Customer Churn dataset in R and found Logistic Regression achieved the best performance, with an accuracy of 82.30% on this dataset.
Abstract
Predicting customer churn is essential for improving retention and supporting long-term business growth. In this study, we compared explainable machine learning models for predicting customer churn using the IBM Telco Customer Churn dataset in R. Our approach included data preprocessing, exploratory analysis, model development, performance evaluation, and further analysis. We developed and evaluated four classification algorithms: Logistic Regression, Decision Tree, Random Forest, and Extreme Gradient Boosting, using an 80:20 train-test split. We assessed each model’s accuracy, precision, recall, and F1-score. Logistic Regression achieved the best performance, with an accuracy of 82.30% on this dataset. Feature importance analysis indicated that contract type, customer tenure, monthly charges, total charges, and internet service were the key factors influencing churn. These results suggest that explainable machine learning offers both strong predictive performance and greater transparency. The R-based framework we present provides a practical, reproducible approach to support customer retention strategies and help managers make evidence-based decisions in customer relationship management.
