CREDIT RISK EVALUATION BASED ON SUPERVISED LEARNING ALGORITHMS

Credit risk evaluation is a very important management science problem in the financial analysis area. The purpose with this paper is to present tools that may help to identify and foresee which clients will be good credit payers (or not) in relation to credit from banks. This paper investigates the impact of eight feature ranking and selection methods on seventeen classifiers on real life credit risk dataset. Consequences of choosing different supervised learning algorithms are monitored, together with the effects of different feature ranking and feature selection methods. Experimental results demonstrate the effectiveness of feature ranking and selection methods for different supervised learning algorithms. Although data in thing problems involve, in general, thousands or even millions of data, different from the problem presented here, the conclusion arrived at can be used as support for larger problems.