Stability Evaluation Of Landslide Based On Machine Learning And System Development

Landslide stability refers to whether the landslide can maintain a stable state under normal working conditions or heavy rain conditions.In recent years,domestic landslide disasters have occurred frequently,accounting for more than 60% of all geological disasters,causing inestimable losses to the national economy.Correct evaluation of the stability of landslides is the prerequisite for landslide prevention.Establishing a rapid evaluation model of landslide stability is of great significance for landslide prevention.This thesis takes Ningnan County of Sichuan Province as the research area,processes the landslide data of Ningnan County from 2008 to 2016,analyzes and calculates the evaluation factors that affect the stability of landslides in the study area and conducts a chi-square test.On this basis,four machine learning methods were used to build a landslide stability evaluation model and model fusion,and the fusion model was applied in Ganluo County.Finally,a landslide stability evaluation system was designed and implemented based on the above research.The main conclusions of this thesis were as follows:(1)Among the landslide stability evaluation factors,the importance of rainfall and slope is the highest,and the aspect and location of houses are the least important.Chisquare test was performed on the 15 evaluation factors,the highest scores were obtained for rainfall and slope,while the aspect and house location scored 0 with two decimal places.In the subsequent modeling process,the two factors of aspect and house position are eliminated.(2)Four algorithms of XGBoost,Random Forest,LightGBM and Logistic Regression were used to model the landslide stability in Ningnan County,the results showed that the XGBoost evaluation model has the best effect.The evaluation accuracy rate was 89%,which was 1%~9% higher than other models,and the AUC value was 0.87,which was higher than other models 0.05~0.13.The genetic algorithm was used to optimize the model.The results showed that the genetic algorithm requires 29.26 seconds shorter than the grid search,and the optimization efficiency was greatly improved.Model fusion for XGBoost,random forest and logistic regression,the evaluation accuracy of the fusion model was 91%,and the AUC value was 0.88.The evaluation effect was better than the single model.(3)The fusion model was applied to the stability evaluation of landslides in Ganluo County,with an accuracy rate of 89%.After calculating and analyzing the stability evaluation factors of Ganluo County,the fusion model was used for evaluation,the evaluation results showed that 86 of 97 landslide data in Ganluo County were correct and 11 were incorrect,which was basically the same as the evaluation accuracy of Ningnan County,indicating that the fusion model performs well in the evaluation of landslide stability in different regions.(4)The operation of the landslide stability evaluation system was in good condition,the functions required for landslide stability evaluation were realized,and the completed work was well visualized.