Early Warning Prevention Of Landslide Disasters Using Data Mining

China has a vast territory and complex terrain.A large number of landslide disasters occur every year,resulting in huge pro perty losses and serious casualties.Therefore,the research on early warning prevention of landslide disasters has important practical significance.Among the m,how to quickly and accurately realize the stable state analysis of potential landslides,the p rediction of landslide deformation and displacement,and the prediction of the impact range of landslides after the occurrence of landslides has important engi neering and academic value for the early warning prevention of landslide disasters.Based on this,this paper adopts the data mining method and based on the support vector machine theory to build a corresponding prediction model for the potential lands lide stable state,potential landslide surface deformation,and potential landslide horizontal moveme nt distance,and optimize the model accordingly.Finally,the model constructed in this study is verified by taking the Ziwang slope in Guizhou Province as an example.The main work is as follows:(1)In this study,to predict slope stability,a support vector machine(SVM)-based model was established.A dataset of 171 slope cases with six indicators(e.g.,unit weight,cohesion strength and friction angle of slope mass,pore pressure ratio,slope angle,and slope height)was used to train an d validate the SVM model.Based on data characteristics,the effects of the ratio of training set to prediction set,data preprocessing method and parameter optimization search method on the accuracy of prediction model are studied.The results show that:(1)the average accuracy of the prediction model in each case is about 80%.The prediction accuracy increases with the increase in the proportion of the training set.The average accuracy of the model reaches 84.39%;(2)The data preprocessing method has little effect o n the prediction accuracy of the model.(3)The optimized parameter search method of " One grid with N sampling average of each grid cell" has advantages in accuracy and speed compared with the conventional search scheme of " N grids with each grid optimization".(4)The optimized model can predict the stable state of the slope in the test set with an accuracy of 92.31%.(2)Taking time series data or previous displacement as the prediction index,and the target displacement as the prediction object,two kinds of slope surface displacement prediction models based on support vector machine were constructed.According to the model characteristics,the effects of the number of training sets,the type of kernel function,and the length of input va riables on the model are studied.The results show that:(1)In the model with time series displacement as the predictor,the number of training sets will affect the prediction accuracy of the model,and the effects of different kernel function models are different,and the minimum average relative error is 2.9%.(2)In the calculation results of the previous displacement as the prediction index,the length of the input variable will affect the prediction accuracy of the model.The length of the input varia ble that achieves the smallest error in different kernel function models is not the same,and the minimum average relative error is3.4%.(3)Through the verification of an actual landslide case,the average relative error of the optimized model prediction results is 1.3%.(3)Aiming at the uncertainty of landslide parameters and the insufficiency of existing models,taking 424 earthquake and rainfall-induced landslides as the research object,a landslide horizontal motion based on Monte Carlo simulation(M C)and support vector regression(SVR)was established.Distance Probabilistic Prediction Models.The influence of the ratio of prediction set to prediction set,the type of kernel function,the value of parameters(c,g)on the prediction performance of the SVR model is discussed,and the MC simulation method is introduced to predict the probability distribution of the horizontal movement distance of the landslide.Prediction schema for the range of horizontal motion distances.Finally,a method to enhance the applicability of the schema is proposed,and the validity and reliability of the prediction schema constructed in this paper are verified by engineering examples.The results show that:(1)the determination coefficient of each condition was greater than 0.825,with the highest one 0.854 was obtained with the condition of training/testing ratio of 7/1 in combination with the RBF kernel function.(2)Increase of training/testing ratio could increase the prediction accuracy;(3)The model with RBF kernel function performed better than model with other kernel functions;(4)The optimization of(c,g)parameters could significantly improve the prediction accuracy;(5)The feasibility and efficiency of the proposed model was demonstrated via a practical case of Zhonghai Village landslide.Finally,taking Ziwang Slope in Guizhou Province as an example,the required slope parameters were obtained through lab oratory tests,field investigation and instrument monitoring,and the model established above was used to pre dict the stable state of the slope,the deformation and displacement of the slope,and the level of the landslide.Movement distance.The results show that the model established in this paper has good accuracy and applicability,and can provide support for the early warning prevention of landslide disasters in all stages.