CUSTOMER CHURN PROBABILITY PREDICTION FOR ECOMMERCE PLATFORMS USING GRADIENT BOOSTING AND BEHAVIORAL TRANSACTION HISTORY

Abstract

Customer churn prediction has become an essential task for businesses operating in highly competitive industries such as e-commerce, telecommunications, banking, and subscription-based platforms. Customer churn refers to the situation where customers stop using a company's services or products. Identifying customers who are likely to leave the platform is extremely important because retaining existing customers is significantly more cost-effective than acquiring new ones. Therefore, organizations increasingly rely on data analytics and machine learning techniques to predict customer behavior and improve customer retention strategies. This project presents a Customer Churn Prediction System for E-Commerce Platforms using Machine Learning, specifically utilizing the Gradient Boosting algorithm. The system analyzes customer behavioral and transactional data to determine whether a customer is likely to churn. Various attributes such as customer age, tenure, usage frequency, support calls, payment delays, subscription type, contract length, total spending, and last interaction are used as input features for prediction. The project follows a complete machine learning pipeline including data preprocessing, exploratory data analysis, feature engineering, model training, and performance evaluation. The dataset used for this study contains thousands of customer records, enabling the model to learn patterns associated with customer behavior. The Gradient Boosting model is trained on the dataset and evaluated using performance metrics such as accuracy, precision, recall, F1-score, and ROC-AUC score.