Evaluation of Landslide Susceptibility in Xishan Coalfield Based on Information Value Method Coupled with Machine Learning Model
The underground goaf formed by coal mining activities are highly susceptible to geological disasters,and landslide susceptibility evaluation is a prior work for geological disaster risk warning.Taking Xishan Coalfield in Shanxi province as the study area,20 landslide disaster-causing factors are constructed,while three coupled machine learning models of Ⅳ-LR,Ⅳ-RF and Ⅳ-SVM are constructed by using Information Value(Ⅳ)method coupled with Logistic Regression(LR),Random Forest(RF)and Support Vector Machine(SVM)models,and then,landslide susceptibility evaluation is conducted in the study area by using Receiver Operating Characteristic(ROC)curve,mean and standard deviation to analyze the accuracy of modeling results.Results show that the high and extremely high susceptibility areas in the study area are mainly distributed within 300 m of the water system,while the extremely low and low susceptibility areas are distributed in the central and western regions.The verification accuracy of Ⅳ-LR,Ⅳ-RF,and Ⅳ-SVM models are 76.67％,74.62％,and 78.57％,respectively,and Area Under Curve(AUC)of the ROC are 0.86,0.83,and 0.84,respectively.The Ⅳ-LR model has the highest AUC value and the highest prediction accuracy.
landslide susceptibilityXishan CoalfieldIV methodmachine learning model
万方数据
维普
