From Segmentation to Prediction: A Unified RFM–Machine Learning Model for Online Retail Analytics

Abstract

Customer analytics features in the digital retail age that provides an understanding of the buying behaviour and makes improvements in marketing practice. As e-commerce platforms have massive volumes of transaction data, predictive analytics can provide an effective customer segmentation and future behavioural prediction methodology in doing business. Most online retailers still do not have a means to identify high-value customers correctly or forecast repeat buy. Even though they have access to detailed transaction logs, they cannot use the information effectively. The lack of this insight can easily lead to ineffective targeting and thus unreached revenue potential. In this paper, the researcher tries to study purchase behaviour segmentation with the help of RFM (Recency, Frequency, Monetary) and employ machine learning capabilities to learn customer purchasing behaviour. Using the Online Retail dataset, this study uses data preprocessing, customer segmentation based on RFM scoring, Logistic Regression and Random Forest models to predict the future purchasing behaviour. In the analysis, the frequency and monetary score of customers provide a strong indication of the ability of customers to make repeat purchases. Of all the classifier models tested, Random Forest was found to be more effective and accurate than the precision of other models confirming its suitability in loud customer tests. The findings are beneficial to e-commerce websites that wish to maximize on the strategy of personalized marketing and management of customers. The opportunity to increase predictive accuracy even more is to address hybrid models and real-time segmentation in the future studies.