Spatial Web Based Evacuation Time Prediction Using XGBoost in Tsunami Prone Regions

This study aims to analyze and predict the tsunami evacuation time in Padang City using an artificial intelligence method, namely Extreme Gradient Boosting (XGBoost) Regressor. The data used included distance to the beach, altitude, population, shelter capacity, and evacuation zone area. Model performance evaluation was carried out by accuracy measurements such as MSE, RMSE, MAE, MAPE, and determination coefficient (R²). The results of the analysis show that the XGBoost Regressor provides better prediction performance than Linear Regression and Random Forest. The XGBoost Regressor model is able to achieve an R² of 0.95, MSE of 0.0156, RMSE of 0.1250, MAE of 0.0212, and MAPE of 0.16%. The most influential factor on the evacuation time is the distance to the coastline.. This research has a uniqueness that lies in combining a machine learning-based predictive model with an interactive web interface that utilizes the Google Maps API, so that users get an informative and easy-to-understand spatial visualization. This application is specifically designed to support quick decision-making in tsunami-prone areas by providing real-time evacuation time estimates as well as spatial visualizations. These findings not only provide a scientific contribution to the development of data-based prediction systems but also practical contributions in the form of application prototypes that can be used by the community and related agencies, such as BPBD, in planning and carrying out evacuations more effectively. Thus, this research is expected to improve the preparedness of coastal communities and strengthen an adaptive and technology-based disaster mitigation system.