FACTOR ANALYSIS OF RESIDENTS EVACUATION CHOICE BEHAVIOR DURING HEAVY RAIN USING EXPLAINABLE MACHINE LEARNING MODEL

Abstract

In recent years, heavy rain disasters have caused human damage due to various factors. Therefore, it is still necessary to implement the issues related to the promotion of residents evacuation. In this study, we constructed a resident evacuation choice behavior model using a machine learning model and analyzed the factors of resident evacuation choice behavior using explainable machine learning model. Specifically, PI analysis and PD analysis of the XAI method were performed on the survey data of Gifu prefecture and western Japan during the heavy rain in July 2018 and eastern Japan of typhoon No. 19 in 2019. As a result, we clarified the factors that influence evacuation / non-evacuation and choice of evacuation site.