Application of HLVQ and G-Prop neural networks to the problem of bankruptcy prediction

Abstract

Predicting the failure of a company is a difficult problem traditionally performed by accounting experts using heuristic rules extracted from experience. In this work we apply HLVQ, a new algorithm to train neural networks, to this problem and compared its results with G-Prop, a neural network optimized with evolutionary algorithms. We show that HLVQ is an efficient alternative for the bankruptcy prediction problem. © Springer-Verlag Berlin Heidelberg 2003.