E-Commerce Customer Churn Prediction

Customer churn prediction is essential for helping businesses retain customers and reduce revenue loss. This study compares four machine learning models Logistic Regression, Random Forest and Gradient Boosting using a synthetic dataset created to reflect typical customer behavior. While the use of synthetic data limits real-world generalization, it enables controlled analysis of key features such as purchase frequency, cart abandonment, and session activity. Among the models tested, Random Forest delivered the strongest performance. Future work will focus on real-world data and advanced generative techniques to further improve accuracy and practical use. The models were evaluated using multiple statistical measures, including ROC–AUC, precision–recall, and confusion-matrix based validation, and the study also highlights the limitations arising from the use of synthetic data, which may affect realworld generalization.