Credit risk assessment based on potential support vector machine

A potential support vector machine based learning approach is proposed in the paper to solve the problem of classifier establishment and feature selection in credit risk evaluation. Firstly, previous researches are argued and investigated based on literature review, with main problems faced by researchers in the domain of credit risk assessment concluded. Secondly, the methodology proposed in the paper is also argued in details based on introductions to potential support vector machine, which is a new machine learning method with some differences to the method based on standard ones. So, a new credit risk assessment model based on potential support vector machine, which can accomplish classifier development and feature selection simultaneously, is put forward in the paper. Moreover, the results of experiments based on UCI dataset illustrate that the proposed method has much better generalization performance and less computation consumptions than other ones based on standard support vector machine or artificial neural network.