The Application Of Data Mining In Prediction And Evaluation Of Landslides And Debris Flow

China is a mountainous country,and there are plenty of geohazards,such as landslide and debris flow.To be honesty,the economic losses caused by the geological disasters(landslide and debris flow)in our country are forefronting all over the world.Recently,Our Chinese have more frequent activities in mountainous hilly region since the rapid development in economical.So we now facing an unprecedented pressure coursed by disaster prevention.Undoubtedly,it is impossible to conduct all-round management of all potentially destructive hazards.To solving the problem with severity and contradictory characteristics,this paper used data mining in prediction and evaluation of landslide and debris flow.For case studies,this paper seleted three typical problems such as landslide displacement prediction(Baishuihe landslide),runout distance estimation and hazards mapping(Heifangtai landslides rigion)and prevention engineering treatments evaluation(wenjia gully)based on available information.The main methods and results are as follow:(1)Landslide displacement prediction modelThe curve landslide cumulative displacement is usually nonlinear.Hence,it is challenging to build predictive models with less error.In this paper,we propose a new methodology of embedding wavelet analysis with basic extreme learning machine(ELM)and online sequential extreme learning machine(OS-ELM)to predict the cumulative displacement.First,by wavelet transformation,the cumulative function of displacement is discretized into periodic displacement and trend displacement.Second,basic ELM and OS-ELM are selected to predict the periodic displacement and trend displacement.At last,the cumulative displacement function is computed by ensembling the predicted periodic and trend displacement values.For basic ELM,a sigmoid function is selected as the kernel function and a single hidden layer with 33 nodes performs best.For OS-ELM,the prediction error reaches its minimum with 100 hidden nodes when the RBF function is selected as the kernel function.RMSE for ELM is 0.1423 and for OS-ELM is 0.1315.This methodology with high predictive accuracy performs better in comparison with other methods.(2)Landslide runout distance estimation and hazards mapping modelHazard mapping construction is a prevailing part of spatial analysis of landslides.Estimation on runout distance of landslide better mitigate potential hazards.Conventional mechanism-related methods require series of experiments and/or numerical simulation that are commonly time-consuming and expensive,yet data-driven models reduce the experimental workload and require less prior knowledge in the geological history as well as mechanical behavior of the material.A data-driven model is proposed to forecast landslide runout distance using geometrical characteristics of the landslide.Geometrical dataset of the shallow loess landslides and loess-bedrock landslides occurred in Heifangtai terrace,China,was employed to develop the model.All geometrical datasets were obtained from field investigation and monitoring.Seven data-mining techniques were used and compared for runout estimation,among which the most optimal technique was integrated in the estimation model for loess slope failures.The multi-layer perceptron method outperforms other algorithms,and thus it was selected for the runout distance estimation model.Parametric models are constructed to fit runout distance based on the estimation.Hazard analysis measurements,including Value-at-Risk(VaR)and Tail-Value-at-Risk(TVaR),are computed for the parametric distributions,which shows the potential area of impact and number of residential clusters at risk.(3)Prevention engineering treatments evaluation modelThere is no intuitive comparison of engineering treatments evaluation between before and after governance in previous studies.Based on this purpose,this paper presents a data-driven method to conduct this research by collecting data during five seasons of heavy rainfalls.Survival analysis,Bootstrap method and Extreme Learning Machine(ELM)are selected to build data-driven models.By implementing survival analysis models,the survival probabilities of locations without treatment decrease to 0%.The survival probability of the locations with treatment stay at 55.6% after the five rainfall seasons.Meanwhile,the maximum hourly rainfall in the post-treatment period is 2.571 times as the one in pre-treatment period.Total rainfall volume of post-treatment period is 1.232 times as the one in pre-treatment period.Total time period of rainfall of post-treatment is 5.435 times as the one of pre-treatment.Comparing the survival probabilities,the effect of treatments are significant in the prevention of debris flows.After resampling by bootstrap method,the predictive results from Extreme Learning Machine indicate that,without geological treatment,the probability of having debris flow under 10 times of heavy rainfall is 100%.It is much higher than the observed 30% between the year of 2011 and 2015.The predicted debris flows magnitudes are also significantly higher than the observed ones with treatment.Hence,geological engineering treatment is crucial in reducing and preventing geohazards.