Personal Credit Scoring Based On Classification Tree And Support Vector Machines

The personal credit scoring is an important content in financial and banking circle, and it is essentially a classification problem. With fast development of economy, the importance of personal credit scoring is being strengthened day by day, and the research of personal credit scoring method is becoming more active. Various statistical and non-statistical approaches have been applied in credit scoring. Among them, the classification tree method which is also called recursive partitioning approach is more suitable for dealing with qualitative variables than quantitative variables; Support vector machine is a new machine learning method based on the Statistical Learning Theory, and it is more suitable for dealing with small sample statistics study. Due to excellent ability of studying and extension, the support vector machine has already become hot point of the classification research in credit scoring.Firstly, this paper introduce the personal credit scoring and it's methods, especially the classification tree and support vector machine method. Above this foundation, this paper set up a synthetic approach which combines the above two methods together for credit scoring: first, the classification tree method is applied for qualitative variables, and then the support vector machine is applied at each terminal node for quantitative variables. It has made up insufficiency of the classification tree method bad processing quantitative variables, and can display the advantages of support vector machine in dealing with small sample statistics study. Finally, empirical test shows that this approach outperforms using the support vector machine and classification tree method alone.