Credit risk analysis using machine learning algorithms

Abstract

In credit risk analysis, besides assessing risk of credit applications, taking decision by foreseeing risk of active credit is very important to decrease risk of financial institutions. In Turkey, recent studies reveal that for financial institutions, risk of SME credits is higher than other credit types such as consumer and corporate. Therefore, this paper focuses on predicting SME customer status for period of six months by utilizing application scoring additional to customer behaviour features. By utilizing Random Forest, Neural Networks, Support Vector Machines and Gradient Boosting, performance comparison and also feature analysis for customer behaviour are conducted. Finally, conducted experiments show that utilizing Stacked Generalization methods has positive effect on performance of SME credit risk analysis.