Evaluation Of Landslide Susceptibility Based On Combination Of Random Forest And Optimal Influencing Factors

The frequent occurrence of landslides in Shuicheng District seriously threatens the safety of residents’ lives and property,and restricts the development of the local economy.To improve its landslide disaster prevention and control capability,through research on landslide disasters and their influencing factors in the area,a landslide susceptibility evaluation model based on computer and geographic information system(GIS)technology was established,which combined multiple models and optimal influencing factors.Quantitative analysis and evaluation were conducted on the susceptibility of landslide disasters in the study area,and suggestions for landslide disaster prevention and control were proposed.The main achievements are as follows:(1)12 environmental factors-elevation,slope,aspect,curvature,engineering geology groups,distance from fault,distance from river system,annual average rainfall,mineral density,distance from road,land use and normalized difference vegetation index(NDVI)-were selected as evaluation indicators by integrating factors such as landform,formation lithology,hydrometeorology and engineering activities.(2)Based on the information value(I),certainty factor(CF)and weight of evidence(WOE),the random forest(RF)algorithm was introduced to construct the I-RF,CF-RF and WOE-RF coupling models to evaluate the susceptibility of landslides in Shuicheng District on the GIS platform respectively.The accuracy of the above models was tested by the receiver operating characteristic(ROC)curve.It can be seen that the prediction accuracy of the coupling model is significantly better than that of the single model,among which the WOE-RF model has the best prediction effect.(3)Further,based on the WOE-RF model,combining the greedy algorithm and the grey relational analysis,a landslide susceptibility evaluation model considering the optimal combination of influencing factors was proposed.According to the ROC curve and the law of susceptibility index,the combination model considering the optimal factors has higher accuracy and lower uncertainty,and the prediction effect is better than the traditional combination model of inducing factors,among which the combination model of the random forest-greedy algorithm has the highest accuracy,the area under the ROC curve(AUC)value has reached 0.851,and the accuracy has increased by 7.18%.Therefore,it has a certain significance for improving the accuracy of the landslide susceptibility evaluation model.(4)The extremely high-and high-susceptibility areas are mainly concentrated in areas with large slopes,broken rock masses,near river systems and intensive human engineering activities.The landslide susceptibility zoning are consistent with the actual situation and can provide a reference for the prevention and control of landslide disasters in this and similar mountainous areas.