Research On The Partition Methods Of Geological Disaster Susceptibility In Fugu County

Geological disasters is a destructive geological event, it often make a serious threat to human life and property and living environment, in order to carry out geological disaster-prone zoning accurately and objectively, foreign scholars have proposed a variety of partitioning methods, but different geological conditions of the region vary widely, the results of different evaluation methods that in the study area zoning evaluation are different, therefore, to find a highest accuracy and suitable geological disaster-prone zoning method is particularly important in the study area.In this paper, basing in geological hazard investigation detailed project, with field surveys, using computer technology, based on the natural geological environment study area which detailed investigation on the analysis of the conditions and characteristics of the development of geological disaster zone, established a research area spatial database. I selected aspect, slope, elevation, lithology, rainfall, fault buffers, river buffers, road buffers, vegetation cover and land-use type as geological disaster-prone area of influence factor, based on GIS software to extract a factor layer. Then, selected information law, logistic regression and support vector machine model polynomial kernel function, Rbf kernel function, S function as the partition method, while taking ROC curve to test the sensitivity of partition method, and finally compared the AUC results, the AUC value of information law is 0.7686, the AUC value of logistic regression is 0.8027, the AUC values of supports three kernel function SVM model are less than 0.8. The results showed that logistic regression method is the most accuracy and reliability partition in the use of geological disaster-prone area of the partition and the most suitable method for the study area, and provides an important reference for similar geological conditions of the disaster-prone counties zoning.