Customer Churn Prediction for Telecommunication Companies using Machine Learning and Ensemble Methods

Abstract

This study investigates customer churn, which is a challenge in the telecommunications sector. Using a dataset of telecom customer churn, multiple classifiers were employed, including Random Forest, LGBM, XGBoost, Logistic Regression, Decision Trees, and a custom ANN model. A rigorous evaluation was conducted deploying cross-validation techniques to capture nuanced customer behavior. The models were optimized by hyperparameter tuning, improving their customer churn prediction results. An ensemble averaging method was also adopted, achieving an accuracy of 0.79 and a recall of 0.72 in the test data, which was slightly lower than that of the LGBM, XGBoost, and Logistic Regression. These findings contribute to the development of more reliable churn prediction models to ameliorate the customer retention rates and the operational performance of the service providers.

Keywords

• customer churn
• XG-Boost
• ensemble method
• logistic regression
• deep learning