A Hybrid Approach for Customer Segmentation and Loyalty Prediction in E-Commerce

Abstract

This study aimed to enhance e-commerce customer segmentation and loyalty prediction by integrating machine learning (ML) with traditional statistical methods.Design/Methodology/Approach : The research adopted a hybrid approach, utilizing k-means clustering for customer segmentation based on recency, frequency, and monetary values, followed by an XGBoost classifier application for loyalty prediction. The methodology involved analyzing actual e-commerce data and comparing results with established industry benchmarks.Findings : The hybrid model demonstrated superior performance over conventional methods, evidenced by improved precision, recall, accuracy, and F1 scores in loyalty prediction, alongside higher silhouette scores and lower Davies–Bouldin indices for customer segmentation.Practical Implications : The approach provided a more generalized, interpretable, and high-quality framework for e-commerce businessesto understand customer behavior and enhance retention strategies.Originality/Value : The research contributed to the field by presenting a novel method that successfully combines ML and statistical analysis, offering a more effective solution for customer segmentation and loyalty prediction in e-commerce settings.