Analysis of Landslide Disaster Susceptibility Based on Information Quantity and Machine Learning:Take the Central and Western Regions of Hunan Province as an Example
This study focuses on the central and western regions of Hunan province,with 15 factors selected from four aspects:topography,geological conditions,environmental conditions,and anthropogenic activities to con-stitute a landslide susceptibility evaluation system.First,using 1 017 historical landslide disaster data from 2015 to 2022 as sample data,the information value model was employed to statistically analyze the importance of landslide factors in the regions.The results indicate that the topographic wetness index,land use type,distance from roads,distance from rivers,and vegetation coverage are significant influencing factors for landslide occurrence.Additional-ly,combining with the factor correlation analysis,14 factors were selected for landslide susceptibility zoning analy-sis.Five machine learning methods were applied to model and compare landslide susceptibility,revealing that the random forest method exhibits relatively high accuracy for the relevant factor system in these regions.Finally,the study employed the random forest method to create a susceptibility zoning map for the studied areas,classifying it into five levels:low,low-to-moderate,moderate,moderate-to-high,and high susceptibility.In comparison with the historical landslide records,the zoning results are generally consistent with the landslide densities.The findings pro-vide technical references for landslide disaster prediction and prevention in the central and western regions of Hunan province.
information content modelmachine learninglandslide susceptibility analysiscentral and west-ern regions of Hunan province
万方数据
