A kind of slope stability evaluation model based on SVM-DS method

摘要

This paper provides a recognition method based on support vector machine and D-S evidence theory to get the state of the slope stability timely and accurately. Firstly, the classification of effective recognition model is established by limited empirical data using support vector machine approach. And then the sigmoid function is used to achieve the posterior probability output, which serves as the basic probability assignment in the D-S evidence theory of the model. Therefore, the categorised results are outputted when all the information of evidence is fused according to D-S theory. So the predicted slope stability model of SVM-DS is achieved. The proposed method is tested on a dataset of known slope. The experiment results confirm that this method can greatly enhance the classification accuracy of slope stability.