A new dynamic credit scoring model based on clustering ensemble

With the rapid development of credit industry, customer credit scoring issue is particularly important. In this paper, a new dynamic credit scoring model based on clustering ensemble is proposed to solve the problem that cannot predict customer credit dynamically as well as population drift in customer credit scoring. Firstly, the training set samples are clustered into multiple subareas using OCA clustering ensemble algorithm to weaken the differences among different subareas samples. Then, the entire observation period is fractionized into several fractional periods. Finally customer credit scoring sub-classifiers are established using cost-sensitive support vector machine. The empirical results show that the dynamic model we proposed not only has lower misclassification rate than static model, but also can predict the bad customers as early as possible.