Business failure prediction using neural networks and wavelet neural networks

Bankruptcy prediction models concerns for decades both academics and practitioners. Moreover, in recent years, during the financial crisis period the development of accurate business failure prediction models is particularly compelling. In this paper, we compare the classification performance of two non-parametric approaches, neural networks vs wavelet neural networks on a sample of 240 Greek companies using the financial ratios of Altman’s Z score model. Our results show that the wavelet network outperforms the classical neural network out-of-sample. Moreover, the wavelet network is able to learn identify both healthy and non-healthy firm. On the other hand, neural networks are biased towards non-healthy companies.