Customer churn represents one of the most critical challenges in the telecommunications industry, where retaining existing customers is significantly more cost-effective than acquiring new ones. This paper presents ChurnGuard AI, a comprehensive end-to-end machine learning-based system designed to predict customer churn using the IBM Telco Customer Churn dataset comprising 7,043 customer records and 21 attributes. The proposed system follows a complete data science lifecycle, including data preprocessing, exploratory data analysis, feature engineering, model development, evaluation, deployment, and monitoring strategy. Four classification algorithms-Logistic Regression, Random Forest, XGBoost, and Support Vector Machine-are implemented and evaluated using fivefold stratified cross-validation. The model selection is based on the F1-score metric to effectively handle class imbalance.The best-performing model is deployed through a Flask-based web application that enables real-time churn prediction. The system provides churn probability, risk classification (Low, Medium, High), key contributing factors, and personalized retention strategies. Additionally, a production monitoring framework is designed to detect data drift and ensure long-term model reliability. The results demonstrate that the proposed system achieves strong predictive performance with practical applicability, making it a valuable decision-support tool for telecom operators to enhance customer retention strategies.
Ujitha et al. (Thu,) studied this question.