International Journal of Innovation Engineering and Technology Development نشریه بین المللی مهندسی نوآوری و توسعه فناوری 2 2 2026 03 08 Predicting the Spatial Resilience of Hospitals Against Natural Hazards Using a GIS-Based Hybrid Fuzzy-Machine Learning Model Predicting the Spatial Resilience of Hospitals Against Natural Hazards Using a GIS-Based Hybrid Fuzzy-Machine Learning Model 44 55 FA Rana Naemi Kermanshahi M.Sc. Student in Industrial Engineering – Systems Optimization, University of Urmia, Rana Naemi Kermanshahi M.Sc. Student in Industrial Engineering – Systems Optimization, University of Urmia,fhfhf Ali Doniavi Associate Professor, Department of Industrial Engineering – Systems Optimization, University of Urmia Ali Doniavi Associate Professor, Department of Industrial Engineering – Systems Optimization, University of Urmia Saeed Jafarzadeh Ghoshchi Associate Professor of Industrial Engineering, Urmia University of Technology Saeed Jafarzadeh Ghoshchi Associate Professor of Industrial Engineering, Urmia University of Technology 2026 02 22 This study predicts and spatially delineates the resilience of hospitals against multiple natural hazards (earthquakes and floods), while simultaneously accounting for emergency accessibility, sustainable solar energy potential, and population demand pressure. The main objective is to provide a quantitative and operational tool for identifying high-risk hospitals and prioritizing resilience-enhancement interventions at the urban scale. The methodology is based on a fully integrated three-stage hybrid framework developed in the ArcGIS Pro environment. In the first stage, twelve thematic layers—including digital elevation model, slope, distance from active faults, precipitation, temperature, land use, solar radiation, road network, population density, building density, hospital locations, and power transmission lines—were weighted using the advanced IVFFN-LOPCOW approach to minimize inherent data uncertainties. In the second stage, five key composite indicators (earthquake risk, flood risk, emergency accessibility, sustainable energy potential, and demand pressure) were modeled using the Random Forest algorithm, resulting in highly accurate continuous maps. In the third stage, the final classification of hospital spatial resilience (five classes from very low to very high) was performed using the Support Vector Machine (SVM) algorithm. Findings show that the predictive accuracy of the model for all indicators exceeds 96% (AUC ≥ 0.96). The results reveal that more than 38% of the assessed hospitals fall within low to very-low resilience zones and require immediate structural, infrastructural, and energy-related interventions. The proposed model, as a precise and generalizable decision-support tool, offers strong potential for application in urban planning and disaster risk management within the healthcare system. This study predicts and spatially delineates the resilience of hospitals against multiple natural hazards (earthquakes and floods), while simultaneously accounting for emergency accessibility, sustainable solar energy potential, and population demand pressure. The main objective is to provide a quantitative and operational tool for identifying high-risk hospitals and prioritizing resilience-enhancement interventions at the urban scale. The methodology is based on a fully integrated three-stage hybrid framework developed in the ArcGIS Pro environment. In the first stage, twelve thematic layers—including digital elevation model, slope, distance from active faults, precipitation, temperature, land use, solar radiation, road network, population density, building density, hospital locations, and power transmission lines—were weighted using the advanced IVFFN-LOPCOW approach to minimize inherent data uncertainties. In the second stage, five key composite indicators (earthquake risk, flood risk, emergency accessibility, sustainable energy potential, and demand pressure) were modeled using the Random Forest algorithm, resulting in highly accurate continuous maps. In the third stage, the final classification of hospital spatial resilience (five classes from very low to very high) was performed using the Support Vector Machine (SVM) algorithm. Findings show that the predictive accuracy of the model for all indicators exceeds 96% (AUC ≥ 0.96). The results reveal that more than 38% of the assessed hospitals fall within low to very-low resilience zones and require immediate structural, infrastructural, and energy-related interventions. The proposed model, as a precise and generalizable decision-support tool, offers strong potential for application in urban planning and disaster risk management within the healthcare system. https://ijietp.iloas.com/downloadfilepdf/255883