Credit risk evaluation by using nearest subspace method

Abstract

Listen

Translate article

Abstract In this paper, a classification method named nearest subspace method is applied for credit risk evaluation. Virtually credit risk evaluation is a very typical classification problem to identify “good” and “bad” creditors. Currently some machine learning technologies, such as support vector machine (SVM), have been discussed widely in credit risk evaluation. But there are many effective classification methods in pattern recognition and artificial intelligence have not been tested for credit evaluation. This paper presents to use nearest subspace classification method, a successful face recognition method, for credit evaluation. The nearest subspace credit evaluation method use the subspaces spanned by the creditors in same class to extend the training set, and the Euclidean distance from a test creditor to the subspace is taken as the similarity measure for classification, then the test creditor belongs to the class of nearest subspace. Experiments on real world credit dataset show that the nearest subspace credit risk evaluation method is a competitive method.