Bankruptcy Prediction: Discriminant Analysis versus Neural Networks

The paper presents a comparison between two different approaches to the problem of bankruptcy prediction: the traditional discriminant analysis method and possible solutions based on neural networks. The performance of a simple mathematical model applied to economic and financial ratios obtained from a small sample of manufacturing companies balance-sheets is compared with the performance of some BPN family neural networks trained with the same set of information. The results obtained from the neural networks suggest some interesting improvements and practical applications which are the object of the second and third section of the experiment programme in progress.