A comprehensive framework for customer retention in E-commerce using machine learning based on churn prediction, customer segmentation, and recommendation

The cost-efficiency of retaining existing customers in the realm of e-commerce, the pursuit of new customer acquisition is no longer considered a prudent strategy. By developing a comprehensive customer analytics platform that combines customer segmentation, recommendation, and churn prediction techniques to counter customer attrition, this work fills a gap in the literature. This framework comprises seven integral components, encompassing data pre-processing, exploratory data analysis, churn prediction, feature ranking, cluster tendency analysis, customer segmentation, and recommendation. Experimental analysis for customer churn prediction shows that CatBoost performs the best in the dataset based on accuracy and F1-score and recursive feature elimination is applied to rank the features for customer segmentation to build clusters. Cluster analysis demonstrates that K-means outperforms Hierarchical Clustering and DBSCAN, with a Hopkins score of 0.09322634776929459 confirming this. Collaborative filtering is more effective than popularity-based approaches for recommendations. With the shortest test time and 89% precision, the SVD model yields an RMSE of 1.26, MAE of 0.99, and MSE of 1.69. With data-driven insights and tailored experiences, this integrated framework shows notable benefits in every phase, improving customer retention.