Applycation Of Support Vector Machine To Personal Finance Customer Market Segmentation

Nowadays, personal wealth of the residents in our country is increasing constantly. The financing needs are also showing diversified and personalized features. Financial services of domestic banks change from the unification and standardization to the tailor made for individuals. Banks introduce the market segmentation concept with which the customer market is divided. On this basis, banks provide a target and selective introduction to financial products for different customer groups. They identify different kinds of clients according to different levels of their risk tolerance, and then develop specific financial products for the clients. This has a beneficial effect on the reduction of product development costs.This research takes the commercial bank client segmentation as the subject, and explores the factors to distinguish investor's risk tolerance style and the selection of segmentation method. First, a detailed analysis of the main factors that affect the style of risk tolerance is made. After the qualitative analysis,some of the factors are selected as the optional index system of risk tolerance style. Base on this, the Personal information scale is designed to collect sample data. While the FinaMetrica financial risk tolerance scale is used to measure investors'risk tolerance style. Then, through the quantitative analysis, the identification index system of risk tolerance type is gained. After comparison of the characteristics and classification accuracy of a variety of segmentation methods, the SVM method is selected. By choosing the appropriate kernel function and using 5 fold cross testing method to determine the penalty parameter and kernel parameter, the classification experiment is completed. The SVM classification model is improved based on the connection of both SVM classification model and KNN classification arithmetic. The improved model is more accurate in classification. It is compared to the classification models of neural network and decision tree, and the results indicate that the improved model has the highest accuracy.It is beneficial to make the client segmentation with the improved SVM model on the basis of the distinguish factors of risk tolerance styles. On one hand, the banks can reduce the financing service cost by developing and marketing the products according to the size of the clients. On the other hand, personal investors can learn their risk tolerance styles clearly, which may provide a guidance for financing product selection.