Customer Churn Prediction in the Telecommunication Industry Using Machine Learning: A Review and Comparative Analysis

Customer Churn Prediction in the Telecommunication Industry Using Machine Learning: A Review and Comparative Analysis

Abstract

Acquiring new subscribers for telecommunications providers is significantly more expensive than keeping existing ones. However, annual churn rate for these providers lies between 15% and 35%. This paper surveys how machine learning has been used to address this problem over the last five years. We examine models ranging from simple logistic regression to gradient boosting ensembles and sequence-aware deep learning models. We compare results on popular benchmarks IBM Telco, SyriaTel, and UCI KDD Orange corpus. We examine accuracy, AUC-ROC, and F1-score metrics. We also examine how these models handle class imbalance using SMOTE. Our survey of over 30 papers indicates gradient boosting models (XGBoost, LightGBM, CatBoost) dominate this domain. CatBoost achieves an AUC of 0.982 on these benchmarks. Explainable AI methods like SHAP and LIME are also discussed both as performance enhancers and regulatory enforcers.

Authors

Nupur Tripathi, Tanay, Ujjwal Ankit, Supriya Singh, Suraj Kumar

Institution

NIET Business School, Greater Noida, India

PDF Link

View Full Paper