Risk Assessment Of Landslide Geological Disasters In Daning County Based On Machine Learning Model

Daning County,Shanxi Province is located on the Loess Plateau in western Shanxi.Its relatively complicated geological environment and intensive human engineering activities have led to the development of many landslide geological disasters in the county.Carrying out a landslide geological hazard risk assessment is a prerequisite for effective prevention and control of landslide hazards.The results of the evaluation can guide the prevention and control of urban geological disasters.Based on the urban geological disaster investigation project in the Luliang Mountain area,this study investigated 208 landslide hazards on the spot and combined the 1:50,000 geological data of Daning County to investigate the development characteristics and geological background of the landslide disaster in the study area in the county Analysis,using artificial neural network model and support vector machine model optimized by hybrid Gaussian clustering model to evaluate the risk of landslide disaster in the study area,compare the effectiveness and accuracy of different models for landslide disaster risk assessment,and finally generate A zoning map of landslide hazard risk in Daning County used to guide disaster prevention and mitigation.The main results of this article are as follows:(1)To systematically elaborate the regional geological background of Daning County,and then to summarize its development characteristics based on the distribution of landslide disaster points in the study area.The landslide hazards in the study area are mainly distributed in the loess area,which has a high consistency with the road distribution,and decreases with the increase of vegetation coverage.(2)Calculate the ratio of the landslide under different influencing factors,the grading ratio and the amount of information to select preliminary index factors,and then use Spearman correlation analysis method to analyze the correlation of each influencing factor,and finally select the elevation,slope,aspect,Ten index factors,such as slope curvature,landform type,rock and soil type,soil erosion,road network density,average annual rainfall,and vegetation coverage normalization index,establish an index system for landslide hazard risk assessment in the study area.(3)Input the amount of information between different factors and the relationship between landslide hazards into the mixed Gaussian clustering(GMM)for clustering,and then select non-landslide units from the low-risk classification of the clustering results.Support vector machine(SVM)and artificial neural network(ANN)models were selected to evaluate the landslide hazard risk in the study area.ROC curve and four machine learning model performance evaluation indicators were used to test the accuracy and generalization of the model.The results show that the performance of the cluster-optimized machine learning model(GMM-ANN-SVM)is better than the individual model.And the optimized artificial neural network model(GMM-ANN)has the best evaluation effect.(4)Use the GMM-ANN model with the best evaluation effect to complete the geological hazard risk mapping in the study area in the GIS software,and divide the risk into five levels of low,low,medium,high and high using the natural discontinuity method,And partition description.