Prediction Of Slope Deformation Based On Improved Grid Search Support Vector Machines Method

The damage caused by slope deformation is very dangerous,it will not only destroy the construction itself,but also can change the geomorphological characteristics of a region,and then have a significant impact on the surrounding human settlements.The losses caused by a slope disaster are often in the tens of millions or even billions,and the cost of preventing the deformation of these slopes is far less than the reconstruction and recovery work after the disaster.Therefore,it is very important and urgent to predict slope deformation scientifically and effectively.According to the characteristics of slope deformation,this paper proposes support vector machine(SVM)method based on improved grid search methods for prediction of slope deformation.The experimental results show that the improved algorithm can improve the operation time and precision better than the two other traditional algorithms.The work of this paper mainly includes the following aspects:1)firstly,this paper outline the basic theory of SVM,and the kinds of kernel function of SVM,then outline the parameter optimization problem of SVM kernel function.2)Related to the kernel function parameter optimization problem has not been solved,this paper emphatically analyse the traditional grid search method and particle swarm optimization method for parameter optimization search,Due to the deficiency of the traditional grid search method,this paper introduce particle swarm optimization(PSO),which can converge the neighborhood of the optimal solution quickly,in the new algorithm,first use PSO for coarse search then use grid method for fine search with small length.To some extent,it helps the algorithm to jump out of the local optimal solution which the particle swarm optimization may fall into,so as to achieve the global optimal solution.3)Finally,the algorithm is applied to the prediction of slope deformation in practical engineering,through two examples of slope deformation to verify the effect of improved algorithm.The results show that compared with two other conventional algorithm,the new algorithm has smaller average relative error and shorter computational time,On the other hand,the predicted mean square error and square sum error of the new algorithm are also much smaller than the other two algorithms in both engineering examples.This proves that the improved algorithm has better prediction accuracy and stability,and meets the needs of engineering practice.