Spatial Heterogeneity of Neighborhood‐Scale Flood Vulnerability in Coastal Megacities: An Integrated Machine Learning– MCDM Framework

ABSTRACT Climate change and rapid urbanization are intensifying flood hazards, jeopardizing urban systems and socio‐economic stability. Robust Flood Vulnerability (FV) assessment is therefore essential for proactive mitigation and adaptation, yet existing approaches remain divergent and fragmented. This study develops a composite FV Index (FVI) based on the conceptual foundations of the Measurement of Vulnerability in Europe (MOVE) and Baseline Resilience Indicators for Communities (BRIC) frameworks. An initial pool of 58 indicators across five dimensions—physical, socio‐demographic, economic, institutional, and environmental—was reduced to 34 via the Kaiser–Meyer–Olkin (KMO) test. Indicator weighting used a hybrid multi‐criteria decision‐making (MCDM) approach combining the Fuzzy Analytic Hierarchy Process (FAHP) with the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), alongside a data‐driven Adaptive Neuro‐Fuzzy Inference System (ANFIS). The framework was applied to 101 sheyakhahs (census administrative units) across five districts in Alexandria, Egypt, a coastal city highly exposed to hydroclimatic stressors, where the FVI was mapped and validated. Results show strong agreement between ANFIS and FAHP–TOPSIS outputs (Pearson's r = 0.91, p < 0.01), and Monte Carlo sensitivity analysis confirms high robustness, with 94.1% of units retaining consistent rankings under ±15% weight perturbations. Spatial autocorrelation revealed significant clustering of high vulnerability along the northwestern coastal corridor (Moran's I = 0.88 for ANFIS; 0.74 for FAHP–TOPSIS), validated against building‐damage records (Spearman's ρ = 0.88, p < 0.001), with El‐Gomrok consistently the most vulnerable district. The framework provides a transferable, neighborhood‐scale tool for targeted infrastructure planning, contributing to the achievement of SDG 11.