Customer Segmentation Using Unsupervised Machine Learning Techniques: A Data Driven Approach

Customer segmentation is a critical component of modern marketing analytics and business intelligence. This paper presents a comparative analysis of three widely used unsupervised machine learning algorithms — K-Means, DBSCAN (Density-Based Spatial Clustering of Applications with Noise), and Agglomerative Hierarchical Clustering — for the purpose of customer segmentation in the retail domain. The UCI Online Retail Dataset is preprocessed and transformed using RFM (Recency, Frequency, Monetary) analysis to extract meaningful behavioral features. Each algorithm is applied to the RFM feature space and evaluated using Silhouette Score, Davies-Bouldin Index, and Calinski-Harabasz Index. Cluster visualizations are generated using Principal Component Analysis (PCA) for dimensionality reduction. Experimental results demonstrate that K-Means achieves superior segmentation quality in terms of Silhouette Score, while DBSCAN effectively identifies outlier customers. This study provides a structured benchmark for selecting appropriate clustering methods in customer analytics and offers actionable insights for personalized marketing strategy formulation.