University of Tehran Press Advances in Industrial Engineering 2783-1744 60 1 2026 06 01 A Distribution Network Design Model Using Data Classification and Fleet Optimization 53 70 103942 10.22059/aie.2025.386390.1927 EN Mohammad Amirahmadi Ph.D. Candidate, Department of Industrial Engineering, Faculty of Engineering, Islamic Azad University, North Tehran Branch, Tehran, Iran. Hamid Esmaili Associate Professor, Department of Industrial Engineering, Faculty of Engineering, Islamic Azad University, North Tehran Branch, Tehran, Iran. 0000-0003-0149-8077 Kia Parsa Assistant Professor, Department of Mathematics, Faculty of Sciences, Islamic Azad University, North Tehran Branch, Tehran, Iran. Amin Mostafaee Associate Professor, Department of Mathematics, Faculty of Sciences, Islamic Azad University, North Tehran Branch, Tehran, Iran. Journal Article 2024 12 02 This study seeks to address gaps in previous research by introducing a comprehensive data-driven distribution network design model. The process begins with an in-depth analysis of customer demand, utilizing unsupervised learning algorithms to gain valuable insights into consumer behavior. This analysis identifies demand levels across different geographical regions and reveals temporal demand patterns. The resulting insights serve as inputs to the distribution network design model. To facilitate effective data classification and analysis, The Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is employed. enabling accurate estimation of customer demand based on innovative parameters. Based on the clustering results, a mixed-integer linear programming model is developed that incorporates facility-location, capacity-planning, product-flow, and fleet-composition decisions. Importantly, during this modeling process, emphasis will be placed not only on optimizing the number, location, and capacity of facilities but also on refining fleet types and their compositions to enhance overall efficiency. The proposed model is solved using CPLEX in GAMS and evaluated through a set of numerical test instances. The results demonstrate that the proposed data-driven model achieves an average profit improvement of 10-15% compared to traditional non-clustered approaches. The model also yields savings in transportation and fleet-related costs. Moreover, its integrated structure enables sensitivity analyses of key parameters and provides useful managerial insights. demonstrating the synergy between data-driven clustering and mathematical optimization for distribution network design. https://aie.ut.ac.ir/article_103942_827401f903a22dee8ae1e1ec99655644.pdf