Landslide Susceptibility Evaluation And Landslide Displacement Prediction Based On Particle Swarm Optimization-Based Support Vector Machine

Landslides are uncontrollable but preventable.The preparation of landslide-prone zoning maps is the first choice for dealing with geological disasters.The monitoring and prediction of landslides is also an inevitable choice for successful disaster avoidance.In this paper,the landslide susceptibility evaluation is completed by extracting and tidying the landslide environmental factors and stimulating factors,as well as the historical landslide cataloging and landslide displacement and other multi-source spatial information,and using the support vector machine method based on particle swarm optimization to complete the landslide susceptibility evaluation.The landslide deformation prediction model realizes the prediction of landslide displacement and proves that it has high evaluation and prediction accuracy.Specifically,it includes the following aspects:(1)Based on the historical landslide data of Wushan section of the Three Gorges reservoir area,the relationship between lithology,water system,rainfall distribution and basic environmental factors such as slope,aspect and slope height and landslide distribution are extracted and analyzed,and the landslide density values are used.Qualitative or quantitative factors were normalized.(2)Because support vector machine(SVM)has good adaptability in practical problems such as small sample,high dimensionality and nonlinearity,it is very suitable for the evaluation and early warning of landslide susceptibility.The Particle Swarm Optimization(PSO)algorithm has a powerful global optimal search capability.Therefore,it is proposed to combine the two to construct a PSO-SVM hybrid technology for the susceptibility evaluation of landslides,and finally complete the Wushan section of the Three Gorges reservoir area.The landslide susceptibility evaluation was carried out,and the accuracy of the landslide density ratio and LAR value was evaluated.The results show that it has good consistency with the historical landslide.(3)Combining the characteristics of the deformation sequence of the landslide monitoring point of Miaodian,according to the time series addition model,it is proposed to use the second moving average method to decompose the total displacement of the landslide into the trend term displacement and the factor term displacement;then use the polynomial model and PSO-respectively.The SVR method simulates the trend term and the displacement related to the excitation factor.Finally,the two are superimposed on the displacement of the predicted monitoring point.The calculation results of the four monitoring points of the Miaodian landslide are calculated and analyzed,and the results show that the prediction accuracy is high.(4)Finally,combined with the requirements of landslide hazard monitoring and early warning system construction,the preliminary design of the database and system was carried out,and the forecasting and evaluation model was integrated into the landslide monitoring and early warning system,and the algorithm processing flow was given,in order to be integrated.Regional evaluation and deformation prediction of landslide hazards.