Assessment of Rainstorm Disaster Comprehensive Loss Based on Bayesian Optimized XGBoost in Hunan Province
Xiang Xin
Wang Liping
Wang Weiguo
Wang Guanlan
Abstract：Based on county-level rainstorm disaster data from 2016 to 2022 in Hunan Province,a comprehensive loss evaluation index is constructed by objective methods,which is then classified into four levels.Thirteen explanatory variables are selected by considering disaster-causing factors,environmental factors,disaster-exposed entities and disaster prevention and mitigation capabilities.The comprehensive loss assessment model is developed using Bayesian optimized XGBoost.The results show that:(1)In the comprehensive loss evaluation index,the population loss factor has the highest weight,followed by hous-ing loss and agricultural damage area.(2)The Bayesian optimized XGBoost model demonstrates a good performance,achieving accuracies of 0.86 and 0.83 for the training and testing samples,respectively.(3)The explanatory variables related to disaster-causing factors are the most crucial factors influencing the identification of loss levels,which have the greatest impact(83.11%)on the assessment of severe loss.Particularly,the increase in both the average process rainfall and maximum process rainfall can raise the probability of a sample being classified as a severe loss level.This model is applied and tested for the rainstorm event in Hunan Province from June 1 to 5,2022.Districts and counties that experienced casualties are classified as the severe loss level.The assessment results align well with the collected dis-aster data.
Keywords：rainstorm disastercomprehensive lossXGBoosttesting and assessment
Publication Date：2025-11-30
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：7( 74-80 )