Customer Segmentation using Machine Learning and Interactive Visualization

Abstract

This study proposes an end-to-end workflow for customer segmentation in e-commerce, aiming to bridge the gap between data science and marketing applications. Using a UK online retail dataset, a preprocessing pipeline incorporating "net value offset" logic was built, with logarithmic transformation and Z-score standardization applied to construct a standardized RFM feature space. K-Means clustering was then employed to group customers into four segments, with the optimal number determined by the Elbow Method and Silhouette Score. To overcome visual overlap in linear dimensionality reduction, t-SNE was used for 2D projection, and an interactive Tableau BI dashboard integrating quintile-based scoring and multi-dimensional filtering was developed to translate algorithmic outputs into interpretable business profiles. The analysis identified four groups—"VIP Customers," "At-Risk Loyalists," "New Customers," and "Hibernating Customers"—revealing a Pareto-consistent pattern where a small core of customers contributes a disproportionate share of revenue. Overall, the study offers an integrated framework to support management in optimizing marketing resource allocation.