Applying Clustering Techniques for Customer Segmentation Based on Shipping Behavior, Cost, and Satisfaction in Logistics Services

Abstract

In modern logistics operations, behavioral data-based customer segmentation plays a crucial role in optimizing service delivery and achieving competitive differentiation. This study proposes a clustering-based approach using K-Means, Agglomerative, and Gaussian Mixture to segment sender-level customer profiles in a logistics network based on shipping cost and delivery duration, while customer satisfaction is used for post-cluster profiling and interpretive analysis. A comprehensive preprocessing pipeline is implemented, including temporal feature engineering and sender-based statistical aggregation. Grid search is used for hyperparameter tuning, and clustering performance is evaluated using the Silhouette Score, Calinski-Harabasz Index, and Davies-Bouldin Index. The results indicate that K-Means with two clusters achieves the highest silhouette score (0.843), outperforming the aggregative and Gaussian mixture models. Principal Component Analysis (PCA) reveals clear separability between clusters labeled as Efficient Senders and Costly & Slow Senders. These findings provide actionable information for logistics service providers to improve pricing strategies, delivery efficiency, and customer satisfaction through intelligent segmentation.