Customer Segmentation Using Machine Learning

Abstract

This study examines how integrating Recency, Frequency, and Monetary (RFM) analysis with unsupervised machine learning improves customer segmentation and churn insight using transactional retail data. RFM variables were computed for 3,000 customers to capture purchase recency, transaction frequency, and spending intensity. Results show strong heterogeneity and a highly skewed monetary distribution, indicating that a small share of customers generates disproportionate revenue. K-means clustering produced six interpretable segments ranging from loyal high-value customers to dormant customers, with clear economic and behavioral separation. Segment-level analysis demonstrates a monotonic increase in churn from high-engagement to low-engagement clusters, with churn rising from 4.2% in the loyal segment to 48.9% in the dormant segment. DBSCAN additionally isolated noisy and irregular purchasing patterns, improving robustness to outliers but reducing managerial simplicity. Overall, the findings support machine learning enabled, value-based segmentation for targeted retention and marketing strategy development.