Scholarlink

<p>Strategic distribution center (DC) location selection is a long-term decision that should simultaneously consider sustainability, resilience, and future uncertainty. However, existing studies rarely integrate these aspects within a unified AI-driven decision-making framework. To address this gap, this study proposes a novel hybrid framework that combines the Goal Programming-based Stochastic Best–Worst Method (GPSBWM), Weighted Fuzzy Inference System (WFIS), Random Forest (RF), and SHAP explainability analysis. GPSBWM is first employed to determine robust criteria weights under multiple future scenarios. Subsequently, WFIS generates reliable labels for previously unlabeled data, enabling the development of a supervised Random Forest classifier. Finally, SHAP is applied to explain the contribution of individual criteria and enhance model transparency. The proposed framework is validated using a real-world case study involving 200 candidate distribution center locations. The results identify security (0.0923), costs (0.0801), accessibility (0.0774), capacity expansion (0.0764), and flexibility (0.0769) as the most influential criteria. Furthermore, the developed framework successfully classifies candidate locations into Selected, Saved, and Rejected categories, with 10 locations identified as the most suitable alternatives for distribution center establishment. Comparative analyses demonstrate the effectiveness and robustness of the proposed approach, while SHAP provides transparent managerial insights into the contribution of each evaluation criterion. The proposed framework offers both methodological and practical contributions for sustainable and resilient distribution center location selection under uncertainty.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Explainable AI-Driven Hybrid Decision-Making Framework for Sustainable and Resilient Facility Location Selection

  • 期刊论文

  • 发表日期：

    2026年7月25日

查看全文
  • Kayvan Korani,
  • Armin Tahmasebinezhad,
  • Mohssen Ghanavati-Nejad,
  • Arash Yousefi

摘要

Strategic distribution center (DC) location selection is a long-term decision that should simultaneously consider sustainability, resilience, and future uncertainty. However, existing studies rarely integrate these aspects within a unified AI-driven decision-making framework. To address this gap, this study proposes a novel hybrid framework that combines the Goal Programming-based Stochastic Best–Worst Method (GPSBWM), Weighted Fuzzy Inference System (WFIS), Random Forest (RF), and SHAP explainability analysis. GPSBWM is first employed to determine robust criteria weights under multiple future scenarios. Subsequently, WFIS generates reliable labels for previously unlabeled data, enabling the development of a supervised Random Forest classifier. Finally, SHAP is applied to explain the contribution of individual criteria and enhance model transparency. The proposed framework is validated using a real-world case study involving 200 candidate distribution center locations. The results identify security (0.0923), costs (0.0801), accessibility (0.0774), capacity expansion (0.0764), and flexibility (0.0769) as the most influential criteria. Furthermore, the developed framework successfully classifies candidate locations into Selected, Saved, and Rejected categories, with 10 locations identified as the most suitable alternatives for distribution center establishment. Comparative analyses demonstrate the effectiveness and robustness of the proposed approach, while SHAP provides transparent managerial insights into the contribution of each evaluation criterion. The proposed framework offers both methodological and practical contributions for sustainable and resilient distribution center location selection under uncertainty.