Research On Landslide Prediction Algorithm Based On Improved Support Vector Machine

Landslides are one of the most dangerous natural hazards in the world and pose a serious threat to humans as well as to society.Landslide monitoring and early warning are the main technical means to avoid this hazard.The thesis relies on a multi-sensor landslide monitoring approach and investigates and implements an improved support vector machine based landslide prediction algorithm.The thesis firstly analyzes the influencing factors of the landslide and determines the main parameters of the landslide monitoring.Secondly,the seagull optimized support vector machine algorithm(SOA-SVM)is used to predict the shear parameters(cohesion and internal friction angle)of the mountain by using the main parameter data of the landslide.According to the predicted value of the shear parameters,the safety factor of the mountain is calculated by the limit equilibrium method,the safety state of the mountain is determined,and the early warning of the landslide is realized;finally,the displacement data generated by the later landslide is used and the support vector machine algorithm based on grid seagull optimization(GS-SOA-SVM)carry out displacement prediction to realize the late warning of landslide.The specific research work mainly includes the following three aspects.(1)Research on the main parameters of landslide monitoring.The thesis proposes a method for screening the main parameters of landslide monitoring based on the improved orthogonal test method,which combines the grey relational analysis method and the analytic hierarchy process(AHP)to analyze the influence of 11 parameters such as water content,earth pressure,temperature,rainfall,and soil density on landslide The orthogonal test results of the correlation degree determine the main parameters of landslide monitoring.The test results show that the correlation of the six parameters of water content,rainfall,earth pressure,pore water pressure,temperature,and vibration accounted for 90.79% of the total 11 landslide parameters,while the other five parameters accounted for 9.21%.Therefore,The influence of these 6 parameters on landslides can replace 11 parameters as the basis for landslide judgment and become the main parameters of landslide monitoring.(2)Stability analysis of landslide based on support vector machine optimized by seagulls.First,compare and study the current mainstream regression prediction algorithms,analyze their advantages and disadvantages,and choose support vector machines as the prediction model.The seagull algorithm(SOA)is used to optimize the regularization coefficient C and the kernel function parameter g of the support vector machine,and a landslide stability analysis method based on the SOA-SVM algorithm is established.The results show that the prediction effect of the SOA-SVM algorithm is significantly improved,which is better than the traditional SVM and PSO-SVM algorithm.The safety factor of the mountain is calculated by the landslide stability analysis method based on the SOA-SVM algorithm,the safety state of the mountain is determined,and the early warning of the landslide is realized.(3)Forecast and analysis of landslide displacement based on grid seagull optimization support vector machine.When the safety factor of the mountain tends to or is less than 1 and the displacement data shows a steady increase,the landslide is in an unstable state.At this time,a late warning of the landslide is required.According to the characteristics of the SOA-SVM algorithm,the grid search algorithm(GS)is used to further improve the SOA-SVM algorithm.Using the cumulative displacement of the first three periods as input and the predicted value of the cumulative displacement in the next period as the output,the landslide displacement prediction model is established through the GS-SOA-SVM algorithm and the displacement is predicted.The results show that the GS-SOA-SVM algorithm achieves the expected effect and predicts the displacement at the next moment more accurately.Realize the late warning of landslide according to the change trend of displacement.