Study On The Prediction Model Of Landslide Geological Disaster Based On Machine Learning

China has a vast territory,complex and diverse geographical environment,and the frequent occurrence of geological disasters has caused great harm to people's lives and property.The Chinese Communist Party and State leaders attach great importance to the prevention and control of geological disasters.In order to strengthen the prevention and control of geological disasters and put the systematic prediction of geological disasters on the agenda,the State Council has set up the disaster emergency management department.The traditional geological disaster prediction system has some shortcomings,such as backward monitoring instruments,less installation,single parameters and simple prediction model,These problems will lead to low forecast accuracy.Taking the landslide geological disaster as an example,the paper combines the selection of the factors affecting the landslide geological disaster in Shanyang county with the relevant machine learning theory,analyzes the accuracy of various prediction models,and establishes the integrated prediction model of Bagging based on XGBoost optimization.The advantage of the model is that it can increase the forecast precision.Based on the monitoring data of 12 monitoring stations in Shanyang County,Shaanxi Province,the paper analyzes the types,scale and distribution characteristics of the disasters on the basis of the topography and geomorphology of the research area,and the main influencing factors of landslide geological hazard are determined.Firstly,the landslide disaster monitoring hardware platform designed by the research group is adopted to collect the data from April 2014 to April 2015 in the research area as the sample data;Secondly,the eight major disaster factors of the landslide are determined through data screening by Kernel Principal Component Analysis,at the same time wavelet and Kalman filtering algorithm are used for multi-sensor data fusion,and effective data are selected as training samples;Then,RBF neural network,Support Vector Machine and Bagging ensemble learning model are established to predict landslide disaster;Finally,through R~2and RMSE indexes to analyze the model precision of 170 sets of test data,the results show that the three algorithms have achieved high prediction accuracy,and the Integrated Algorithm is significantly better than the prediction effect of a single weak classifier model.In order to further improve the accuracy of the prediction model,XGBoost is used to optimize the parameters of the integrated algorithm.The results show that the optimized Integrated Algorithm could improve the prediction accuracy of the model.The R~2 index is above 0.9,and the prediction accuracy is improved by 10%.The model improves the reference for the prediction model and ensures the prediction accuracy of landslide disaster.It has certain theoretical and practical significance.There are 45 figures,13 tables and 69 references in the paper.