Abstract:

How to accurately identify and assess stability of landslides is always a key problem in the study of landslides. Based on multi-classification support vector machine theory, multi-classification support vector machine model for landslides stability evaluation was built, by using 37 typical landslides(27 training samples and 10 testing samples ) in the Three Gorges reservoir areas, and was compared with distance discrimination analysis method. The results indicates that the accuracy rates of the SVM model for testing samples and training samples are up to 100%, while the accuracy rates of the distance discrimination method for testing samples and training samples are separately 80% and 77.8%. The identification precision of the former is obviously better than that of the latter. On this basis, the SVM model was applied in the stability identification of Niugundang landslide in the Xiluodu reservoir areas, and the obtained results was in good agreement with the actual situation. So, SVM method has good applicability and effectivity in the stability discrimination practice of landslides and can provide a new way for the discrimination and evaluation of landslide stability.

Key words: multi-classification support vector machine, landslides, stability discrimination, discrimination indexes