Svm-based Bank Customers' Personal Credit Assessment Study

After China joined in WTO, Services of Bank was the main way which bank can benefit from. All those services are not suit all the customers, so the information of customers must be analyzed and sort, and the corresponding customers can be found out. Pass this way, our Bank business could save plentiful cost and improve profitable ability. Even more, our Bank business could gain winning in the stinging competitive international surroundings. With the development of economy and Bank businesses improve rapidly in our country recently, bank gains abounding customers and abundant profit, but it also increases workload in credit evaluation. Credit evaluation model only depending on individual experience in the past is unfit to be used in credit evaluation of bank, so the intellectualized research in individual credit evaluation is more important nowadays.Data mining is a new technique which integrates techniques of database, artificial intelligence and statistic learning. It 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 in data mining. It has excellent theory foundations, which are structure risk minimization, conditional optimization theory and kernel space theory. In order to solve a complicated classification task, the core idea is that SVM maps the vectors from input space to feature space in which a linear separating hyperplane is structured. SVM has the advantage of global optimization, simple structure and good generalization. Recently, SVM has been successfully applied to handwritten digit recognition, face detection and face recognition, text classification fields.Firstly, the paper introduces the application background, Credit evaluation state and credit evaluation method. Bring forward an improved individual credit evaluation method.Secondly, the paper studies SVM theory and algorithm in depth. Research on Platt's sequential minimal optimization algorithm. Implement Keerthi's improved SMO algorithm and linear Kernel, polynomial kernel, RBF kernel and sigmoid kernel support vector machines.Thirdly, customers' credit risk evaluation prototype system based on SVM is designed and realized. Apply the system to customers' credit risk evaluation of foreign Bank. The System implement data preprocessing and SVM classification algorithm (two-class SVM, multi-class SVM), Through prediction accuracy comparison of different models, the paper proves that the improved credit risk evaluation method is better than the old method.Finally, the paper summarizes the whole study work and makes further prospects to the work.