PREDICTIVE ANALYTICS FOR CUSTOMER CHURN IN SUBSCRIPTION-BASED BUSINESSES USING MACHINE LEARNING

Abstract

This research paper focuses on developing a robust framework for predicting customer churn in subscription-based industries. Using the uploaded thesis data, we implemented various machine learning algorithms, including Logistic Regression, Random Forest, and Gradient Boosting. The study emphasizes the importance of data preprocessing and feature engineering. Our findings indicate that ensemble methods provide the highest predictive performance with an accuracy of 88.7% and a superior ROC-AUC score of 0.95. The analysis further highlights that 'Customer Age,' 'Active Membership Status,' and 'Number of Products' are the most significant predictors of churn. This proposed system provides a scalable and adaptive early-warning framework, enabling businesses to implement proactive, data-driven retention strategies to maximize customer lifetime value.