Integrated Fuzzy ANP–Machine Learning Framework with Composite Risk Index for Strategic Passive Defence Prioritization in Gas Infrastructure

Abstract Gas infrastructure systems face escalating threats from both natural disasters and intentional acts such as sabotage and terrorism. This study presents a novel hybrid framework that integrates the fuzzy analytic network process (F-ANP), supervised machine learning, and a composite risk index to systematically evaluate and prioritize passive defense risks in gas facilities. Fifty-six subindicators across eight main criteria were identified through expert judgment and weighed using the Super Decisions software. Fuzzy membership functions were applied to categorize risk levels linguistically. To validate prioritization, random forest and support vector machine models were trained on labeled data from 120 gas sites, achieving cross-validated accuracy exceeding 89%. A composite risk index was then calculated by combining ANP-derived weights, fuzzy scores, and machine learning–based feature importances. Sensitivity and fluctuation analyses confirmed the model’s robustness. The results consistently highlighted emergency response accessibility, structural resilience, and population exposure as top risk indicators. A real-world case study on Facility G-41 demonstrated the model’s operational applicability. The proposed framework enables scalable, interpretable, and evidence-based risk prioritization and addresses key limitations of traditional single-method approaches by integrating both expert insight and empirical data. Future work should incorporate real-time data and broader regional inputs to enhance generalizability.