Applied Density-Based Clustering Techniques for Classifying High-Risk Customers: A Case Study of Commercial Banks in Vietnam

Abstract

Understanding and effectively engaging with customers is paramount in today's rapidly evolving business landscape. With rapid technological advances, banks have unprecedented opportunities to improve their approach to customer segmentation. This change is driven by integrating resource planning systems and digital tools, enabling a more comprehensive and data-driven understanding of customer behavior. Therefore, the study aims to evaluate the performance of various density-based clustering algorithms in classifying customers at risk of default. The algorithms analyzed include K-Means, DBSCAN, HDBSCAN, and Birch, each offering unique strengths in handling diverse data structures. Using a dataset of 77,272 customers from Vietnamese commercial banks spanning 2010 to 2022, the study rigorously assesses these models based on seven critical metrics: Davies-Bouldin Index, Silhouette Score, Adjusted Rand Index, Homogeneity, Completeness, V-Measure, and Accuracy. The results indicate that density-based methods, particularly DBSCAN and HDBSCAN, excel in identifying high-risk clusters despite challenges in cluster separation and alignment with accurate data distributions. Birch demonstrates superior cluster separation and compactness but requires further refinement for optimal accuracy. The findings underscore the potential of integrating clustering methods into credit risk management frameworks, enhancing financial institutions' predictive accuracy and operational efficiency. This research contributes to the ongoing discourse on practical credit risk assessment tools, providing valuable insights for practitioners in the banking sector. Finally, once segments are identified, banks can tailor marketing messages, product offerings, and customer experiences to better suit each group. This can lead to reduced risk, improved customer satisfaction, higher conversion rates, and ultimately increased revenue and customer segmentation in the context of technology trends is becoming an indispensable part of modern business strategy