Distance Prediction of Slope-Foot Landslide in Southwest of China Based on GA-BP Neural Network

Abstract

Landslide is a typical geological hazard in southwest China. In order to reasonably assess the disaster-causing range of earthquake-induced landslides in southwest China, the prediction of landslide movement distance is very important. By making principal component analysis of the factors affecting the distance of landslide movement and combining with the mechanism of landslide movement, the three principal components are named as kinetic energy factor, slope factor and resistance factor, which are used as input neurons of BP neural network. The initial weights and thresholds of BP neural network are optimized by using the global search ability of genetic algorithm, and the prediction models of maximum vertical motion distance (H) and maximum horizontal motion distance (L) of landslide are constructed based on the optimized network. At the same time, on the basis of principal component analysis, a multivariate regression prediction model is constructed, and the results of network model prediction are compared with those of multivariate regression prediction. The optimization effect of genetic algorithm for BP neural network is obvious. The BP neural network model after optimization is accurate and stable in predicting the motion distance of slope foot landslide. The prediction errors of maximum horizontal and vertical motion distances below 10% accounted for 86.67% and 93.33% respectively, which is superior to the original BP neural network and multiple regression prediction model.