Ensemble Learning Approaches for Predicting Flood Damage: A Comprehensive Analysis and Evaluation

This research aims to address the urgent need for a highly accurate predictive model that can estimate postflood damages using historical data of flood events in the United States since January 1950. Floods, as one of the most destructive natural disasters, inflict significant harm on infrastructure, homes, and human lives. Anticipating flood damage is essential for effective disaster management and risk mitigation. Employing the capabilities of ensemble learning techniques such as Random Forest and XGBoost, the model aims to provide near-precise predictions of the potential damage resulting from a flooding event. Simultaneously, the socio-economic dimension of the model integrates demographic data, infrastructure details, and historical economic indicators to comprehend the susceptibility of communities to flood damage.