E-Commerce Customer Segmentation Using K-Means Algorithm Based on Purchasing Characteristics and Customer Satisfaction

Abstract

This study addressed the existing research gap in e-commerce customer segmentation by integrating both behavioral metrics, specifically purchasing characteristics, and emotional metrics, which represented customer satisfaction levels. The primary objective was to establish a highly granular and representative customer typology that traditional transaction-only models fail to capture. To achieve this, a quantitative data mining approach was implemented using a dataset of 450 customer records, which underwent a crucial preprocessing phase using Min-Max Normalization to balance heterogeneous value ranges. The optimal number of clusters was determined using the Elbow Method, and the segmentation was executed through the K-Means Clustering algorithm. The empirical findings revealed that the dataset successfully partitioned into three distinct, non-overlapping behavioral archetypes: Cluster 0 representing high-intensity transactional users with a satisfaction gap, Cluster 1 representing at-risk or dissatisfied customers, and Cluster 2 representing satisfied advocates. The mathematical reliability and strong cohesion of these clusters were rigorously verified by a robust Silhouette Coefficient of 0.62. Ultimately, this research concluded that data-normalized pipelines could successfully transform raw customer analytics into actionable Customer Relationship Management (CRM) strategies, thereby providing e-commerce companies with a reliable foundation to optimize marketing efficiency, mitigate churn risks, and enhance overall profitability.