Susceptibility of debris flow disaster in Hengyang City of Hunan Province based on information value-machine learning coupling model

Taking Hengyang City of Hunan Province as the study area, and selecting the evaluation factors through correlation analysis combined with geographic detector, this paper establishes information value(IV) model with 8 evaluation factors, then constructs the coupling models with random forest(RF) algorithm and light gradient boosting machine(LightGBM) algorithm, respectively, and finally compares the five evaluation models of IV, RF, LightGBM, RF-IV and LightGBM-IV with one another. The ROC curves show that RF-IV model has the highest prediction accuracy compared with the other four models. By comparing the evaluation results of the model with the density distribution data of historical debris flow disaster sites, it is found that the area of high and extremely high susceptible areas accounts for 63.33% of the total. The evaluation results are basically consistent with the actual distribution of debris flow disasters, which can provide reference for subsequent study of debris flow susceptibility as well as prevention and monitoring of debris flow disasters in Hengyang City.