<![CDATA[Customer segmentation based on transactional data is widely used to understand purchasing behavior in retail. However, many existing studies tend to emphasize algorithm performance, with limited discussion on how transactional variables represent actual customer behavior. This study adopts a quantitative approach using transactional sales data from a grocery retail store (Toko Solo Latri), consisting of 10,000 item-level records collected during June 2025. The analysis follows the CRISP-DM framework, covering data understanding, preparation, modeling, and evaluation stages. Customer behavior is represented through several aggregated variables, including transaction frequency, total items purchased, and product diversity. The K-Means clustering algorithm is applied to group customers into meaningful segments. The number of clusters is determined using the Elbow Method and further evaluated using Silhouette analysis. The results reveal three distinct customer segments with different levels of purchase intensity and product diversity. The Silhouette Score of 0.464 indicates a moderate clustering structure. In addition, one-way ANOVA shows significant differences across the observed variables, with large effect sizes (η² ranging from 0.736 to 0.822). These findings suggest that constructing behavior-based transactional features can improve the interpretability of customer segmentation results.]]>
Copyrights © 2026
