International Journal of Science, Engineering and TechnologyAuthors: Dr. Pankaj Malik, Vanshika Soni, Ishaan Nalge, Jatin Bheniya, Ekanki Shrivastava
Abstract: Natural disasters such as earthquakes, floods, and cyclones pose significant risks to buildings, especially in rapidly urbanizing regions. Traditional design methods often fail to capture the complex, multi-factor relationships between environmental hazards and structural vulnerabilities. This study presents a machine learning–driven predictive analytics framework for disaster-resilient building design. The proposed system integrates geospatial information, material properties, structural parameters, soil conditions, and historical disaster data to assess building vulnerability across multiple hazard scenarios. A dataset of 25,000 building samples was used to train and evaluate several machine learning models, including Random Forest, Gradient Boosting (XGBoost), Support Vector Machines, and Artificial Neural Networks. Experimental results demonstrate that XGBoost achieved the highest predictive performance, with 94.5% accuracy, 0.93 precision, 0.94 recall, and an AUC of 0.95, outperforming all other models. Feature importance analysis revealed that soil type, building height, foundation depth, elevation, and reinforcement quality were the top contributors to disaster vulnerability. Model predictions enabled the generation of optimized, hazard-specific design recommendations, improving structural resilience by up to 38% compared to baseline building designs. The findings confirm that machine learning–based predictive analytics can significantly enhance early-stage building safety assessment, making it a powerful tool for architects, engineers, and policymakers in developing disaster-resilient infrastructure
