Application Research Of Housing Loan Credit Evaluation Based On Rough Set And Support Vector Machine

Credit risk is the primary source of risk to financial institutions. With the development of housing marketing in our country, it's very important for commercial bank to establish an effective risk prevention system. Paper is based on DW&MIS platform of China Construction Bank Dalian Branch. Research on individual housing loan credit risk evaluation in enterprise risk management module of DW&MIS platform and implement a credit risk evaluation solution based on Rough Set(RS) and Support Vector Machine(SVM).Data Mining is a new technique which aims at extracting novel and useful knowledge from large volumes of data. Classification is used to predict unknown label by the classifier which is trained with experiential data. It is a basic problem in data mining.Support Vector Machine is a new method based on the idea of VC dimension and Statistical Learning Theory in data mining. The core idea is that SVM maps the vectors from input space to feature space in which a linear separating hyperplane is structured. It is a good classifier to solve binary classification problem and the learning results possess stronger robustness. In this paper we give a hybrid algorithm based on attribute reduction of RS and classification principles of SVM. Finally, the experiment show the effectiveness of the suggested hybrid method, improve the speed and accuracy of prediction.Firstly, the paper introduces the content and problem, individual housing loan state and individual credit evaluation method. Bring forward an improved individual credit evaluation method. Introduce the common classfic algorithms in data mining. By comparing their excellence and disadvantage, giving the reason of using SVM.Secondly, the paper studies SVM theory and algorithm in depth. Comparing three train algorithms, choice Platt's Sequential Minimal Optimization algorithm. Research attribute reduction algorithm of RS, finally choice an algorithm based on significance of attributes.Thirdly, extract classification mining data from DW&MIS platform and preprocess data. Resolve the problem of the application of SVM in credit evaluation, such as the choice of kernel function and parameters, the problem of unbalance data. The experiment show: RBF model is suitable better for practical application, grid-search method adjusts these penalty parameters to achieve better generalization performances in the application, respectively penalty SVM(RP_SVM) can resolve unbalance problem efficiency.