Capacity of Neural Networks and Discriminant Analysis in Classifying Potential Debtors

Abstract

Identifying potential healthy and unsound customers is an important task. The reduction of loans granted to companies of questionable credibility can influence banks’ performance. A prior identification of factors that affect the condition of companies is a vital element. Among the most commonly used methods we can enumerate discriminant analysis (DA), scoring methods, neural networks (NN), etc. This paper investigates the use of different structure NN and DA in the process of the classification of banks’ potential clients. The results of those different methods are juxtaposed and their performance compared.

Keywords

• credit risk
• default
• neural networks
• discriminant analysis
• financial indices
• g33
• c38
• c49