A New Credit Scoring Model Based on Prediction of Signal on Graph

Abstract

Study on high-precision credit scoring model has become a challenging issue. In this paper, a new credit scoring model is proposed. We model the users and their relationship as a weighted undirected graph, and then the credit scoring problem is reduced to a prediction problem of signal on graph. The new model utilizes both the information of the unlabeled samples and the location information of the samples in the feature space, and thus achieves an excellent predictive performance. The experimental results on the open UCI German credit dataset are compared with those of seven classical models that the prediction performance of the proposed model is significantly better than that of the reference models. The Friedman test indicate that the experimental results have a high confidence level.