Abstract
This study focuses on enhancing disaster preparedness in urban areas, with a specific emphasis on Chiba, Japan, following Typhoon Hagibis, through the development of robust flood prediction models. The research investigates the utilization of Artificial Neural Networks (ANN) in flood prediction to enhance disaster preparedness and logistical operations within humanitarian logistics frameworks. By integrating Decision Tree, Random Forest, diffusion, and ANN models, the study aims to determine the most effective approach for flood prediction in urban areas. The research objectives include evaluating the efficacy of different models in predicting flood events by analyzing key parameters such as precipitation, wind speed, and river water levels. Performance evaluation metrics including Correlation Coefficient (CC), Nash-Sutcliffe Efficiency (NSE), Root Mean Squared Error (RMSE), Peak Value, and Percent Bias (PI) will be utilized to compare and assess the outcomes of the prediction models. Results show that the Random Forest model demonstrates superior performance with perfect correlation, efficiency, and minimal errors. The ANN model shows good performance. This research contributes to the enhancement of flood prediction models, disaster preparedness, and response strategies in urban areas affected by natural disasters in Chiba, Japan.