Application of K-means Clustering Algorithm on Customer Segmentation for Effective Marketing Strategy

Abstract

This study discusses the application of K-Means Clustering algorithm in customer segmentation to develop more effective and personalized marketing strategies. In the competitive digital era, companies require deep understanding of customer characteristics to design targeted campaigns. This study uses a dataset of 2,627 customers with 10 demographic and behavioral variables, including age, marital status, education level, work experience, spending score, and family size. The research methodology includes comprehensive data preprocessing (handling missing values, encoding categorical variables, and feature normalization), determining optimal clusters using Elbow method and Silhouette Score, and implementing the K-Means algorithm. The analysis results identified four distinct customer segments: (1) Cluster 0 - premium customers with high purchasing power (12.9%, spending score 2.16), (2) Cluster 1 - mainstream segment with moderate financial stability (47.6%, spending score 1.91), (3) Cluster 2 - large families with moderate spending (22.9%, spending score 1.00), and (4) Cluster 3 - young consumers with growth potential (22.6%, spending score 1.02). Evaluation using Silhouette Score (0.174), Davies-Bouldin Index, and Calinski-Harabasz Index demonstrates good segmentation quality with clear cluster separation. Each segment has unique characteristics that enable the development of specific marketing strategies, such as digital campaigns for young segments, loyalty programs for mainstream market, and premium offerings for high-value customers. This research contributes to the development of data-driven marketing strategies and provides a practical framework for implementing customer segmentation using machine learning in the Indonesian business context.