Analysis Of Landslide In Chongqing City Based On Random Forest Model

The frequency of landslide is higher, harm one of serious geological disasters, cause great economic losses and casualties, does great harm to society. Landslide disasters are widely distributed and occurs frequently in our country, especially in China’s second ladder leading edge, chongqing area terrain fragmented, complex geological structure, rainfall is abundant, along with the development of the western region and the establishment of the three gorges reservoir area, the human engineering activities more and more frequently, makes the regional landslide geological disaster is becoming more and more frequent. Therefore, in order to provide powerful basis of landslide disaster, landslide causes analysis and prediction research area has very important theoretical significance and practical significance. The concrete research content is as follows:(1) this article simply introduces the effects of landslide disaster, and the research significance of landslide, for landslide prediction research both at home and abroad;(2) based on chongqing city from 2000 to 2010 of landslide disaster history data, through the analysis of chongqing city topography slope, slope direction, elevation, and rainfall, selection of rainfall from 2000 to 2010 monthly rainfall, briefly expounds the impact of these indicators for landslide;(3) by comparing the decision tree algorithm, bagging, and the advantages and disadvantages between the SVM algorithm, found that the increasing scale of random forest algorithm with data and data attribute the increase of latitude, able to quickly complete classification prediction, colleges and universities. Therefore, this article finally select random forest algorithm landslide prediction model is set up, try for landslide prediction can be more accurate. Random forests in R language modeling process for the realization of the tool;Research results show that the model of training samples and testing samples are above 80% accuracy, at the same time, the model for the importance of the characteristics of evaluation, found that has the greatest effect in the slope for the occurrence of landslide disaster, slope landslide in between 15 to 25 degrees easily happened, the precipitation is also induced landslide disaster can not be ignored, the important factors that affected by heavy rainfall seasonal activities, the landslide has obvious seasonal, tend to occur in the July to October every year.