An application of SVM-based Classification in Landslide Stability

The calculation method of landslide stability is a critical issue in landslide research. SVM-based multi-classification algorithm, which can structure multiple binary classifiers to accomplish the multi-classification task is used for landslide stability analysis. In this paper, the slope height, slope angle, capacity, internal friction angle and cohesion are selected as impact factors affecting the stability of landslide. Loop crossover method is used to verify the accuracy of the algorithm. Compared with the Mahalanobis distance and Bayes discriminant, the proposed algorithm has a better prediction result, but it also has the largest mis-judgment loss. The accuracy of Bayes discriminant is less than the SVM, but its mis-judgment loss is minimal.