Predictive Vs Descriptive Segmentation: Markov and K-Means Clustering for RFM Analysis

Customer segmentation is an essential analytical strategy that enables businesses to classify their clientele and customize marketing strategies. This study conducts a comparative analysis of two clustering methods-K-Means, which is a descriptive, centroid-based technique, and Markov Clustering (MCL), a predictive, graph-based method-applied to an e-commerce dataset through Recency, Frequency, and Monetary (RFM) analysis. Unlike previous studies that predominantly use linear segmentation models, this research employs graph-based Markov Clustering to identify nonlinear, densely connected customer groups. The dataset underwent thorough preprocessing, including log-normalization and scaling, to ensure consistent feature distribution. Clustering effectiveness was assessed using internal metrics like the Silhouette Score and Davies-Bouldin Index, with MCL showing superior segmentation quality. The findings underscore MCL's capability in identifying subtle customer behavior patterns often missed by K-Means, establishing it as a potent tool for precision marketing and improving customer retention in intricate ecommerce networks.