Skip to Main Content
Credit risk assessment is a crucial area to commercial banks. It's important for banks to discriminate good creditors from bad ones. Support vector machine (SVM) has been applied to classification widely. However, if the index of the training data has much noise and redundancy, the generalized performance of SVM will be weakened, so this can cause some disadvantages of slow convergence speed and low classification accuracy. A SVM classification model based on principal component analysis (PCA-SVM) is presented in this paper, using principal component analysis to reduce the dimensionality of indexes, and then extract principal components to replace the original indexes, and both processing speed and classification accuracy will be improved. At last, apply this model to credit risk assessment, and it shows more generalized performance and better classification accuracy compared with the method of single SVM and BP neural networks.