Ort Vector Machine And Personal Credit Evaluation

Along with the rapid development of consumer credit in domestic commercial banks in our country, personal credit evaluation has been paid unprecedented attention. Scientific and effective personal credit evaluation method has become critical to risk control and to further promote the development of consumer credit for commercial bank. So the research in personal credit evaluation has its important application and practical value. In this paper, on the basis of predecessors’ research, three kinds of personal credit evaluation methods are proposed.SVM has been applied in personal credit evaluation problem for a long time. But the SVM is not necessarily the best choice for bank risk control. So based on partial least-squares regression, this paper poses the first kind of personal credit evaluation method:present partial least squares regression classification method by making activation function. It can control the error rate of the error of the first kind and the error of the second kind by threshold value. The experimental results show that, the present partial least squares regression classification method is feasible, simple and effective.When training data has big modulus, high dimension, and a lot of noise points, the SVM will result in slow convergence speed and low accuracy. So this paper uses the partial least squares regression to reduce the dimensions of the data and uses the components instead of the original data set for training. The second kind of personal credit evaluation method:PLS-SVM algorithm is proposed. The PLS-SVM is applied in personal credit evaluation.In contrast to the SVM and PCA-SVM, experimental results show that the PLS-SVM has clear enhancement in speed and classification accuracy.Different from the SVM, the SVDD algorithm is an unsupervised learning method. It does not need to know the class of samples in the process of learning. So when customers don’t know their class, the SVDD is a good choice for personal credit evaluation. However, unsupervised learning method often has low accuracy, the SVDD is no exception. Finally, by introduced the second-order relaxation variable in standard SVDD, an improved SVDD is proposed and applied in personal credit evaluation. In contrast to standard SVDD and SVM, experimental results show that the improved SVDD has clear enhancement in classification accuracy.