A New Dynamic Credit Scoring Model Based on the Objective Cluster Analysis

摘要

Importance of early prediction of bad creditors has been increasing extensively. In order to solve the problem that cannot dynamically predict customer credit and population drift, this paper presents a new dynamic credit scoring model. Using the objective cluster analysis (OCA) method, the training set is divided into multiple subareas, and the observation period is divided into several periods. Then, customer credit scoring subclassifiers are established using costsensitive support vector machine. The empirical results show that this new dynamic credit scoring model will effectively decrease misclassification rate, and increase accurate rate for bad customers prediction.