AI-Enhanced Disaster Risk Prediction with Explainable SHAP Analysis: A Multi-Class Classification Approach Using XGBoost

Natural disasters pose significant threats to global communities, necessitating advanced predictive frameworks for effective risk assessment and management. This study presents an AI-driven disaster risk prediction system integrating XGBoost machine learning with SHAP (SHapley Additive exPlanations) interpretability analysis. Using the World Risk Index dataset spanning 11 years across 181 countries, we developed multi-class classification models for four key risk indicators: World Risk Index (WRI), Exposure, Vulnerability, and Susceptibility. The XGBoost classifier achieved test accuracies exceeding 0.85 across all categories, with macro-averaged AUC scores ranging from 0.92 to 0.96. SHAP analysis revealed critical driving factors influencing disaster susceptibility, demonstrating the interpretability of AI-powered predictions. Our explainable AI framework provides transparent, actionable insights for policymakers and disaster management authorities, bridging the gap between predictive accuracy and decision-making transparency in global risk assessment.