University of Qom Journal of Data Analytics and Intelligent Decision-making 3115-8161 2 3 2026 07 01 A Hybrid Approach Based on Data Density Analysis for Group Supplier Selection Problem Considering Multi-Level Risk Assessment: A Case Study 1 30 4823 10.22091/jdaid.2026.16147.1066 EN Mohammad Reza Marjani Department of Industrial Engineering , Qom University of Technology, Qom, Iran 0000-0002-7781-2651 Mohammad Habibi Department of Industrial Engineering, Qom University of Technology, Qom, Iran 0009000619601246 Mehdi Jabbari Department of Computer Engineering, Qom University of Technology, Qom, Iran 0009-0008-7674-4408 Majid Aghaee MSc, Faculty of Electrical and Computer Engineering, Qom University of Technology, Qom, Iran. Email: aghaee@qut.ac.ir Journal Article 2026 05 27 Abstract: Supplier selection can be considered as the most important activity in the purchase process because it significantly affects the organization. Despite the presence of powerful suppliers, a variety of risks can be thought of in front of organizations. Thus, it is necessary to identify the related risks, and important decisions should be made by the organization concerning the suppliers with different levels of risks. The problem under consideration, is the group Suppliers selection based on different levels of related risk Criteria. The multi-stage method starts with identifying the risk and after determining the final risk Criteria, the weights of the Criteria are determined using the BWM method. Weighted matrix based on identified quantitative and qualitative risks commensurate with suppliers information and weights of the Criteria, will be used as input to the DBSCAN clustering process. DBSCAN is an unsupervised approach and used here for the supplier selection problem for the first time. We considered the Mega Motor Company, which is the largest manufacturer of powertrains in the Middle East and deals with hundreds of domestic and foreign suppliers. Clustering was made for different values of ε and Minpts. The inter- and intra-cluster distances were used to evaluate the results and clustering quality. Then, the k-means method and unsupervised machine learning were used to assess the validity of the proposed approach. A comparison of inter- and intra-cluster distances of the two approaches verified the efficiency of the DBSCAN approach in dealing with the supplier selection problem. https://jdaid.qom.ac.ir/article_4823_5540e005031b6acb327db8404164219b.pdf