Abstract
This study investigated the use of neural networks as a tool for predicting company failure. The predictive power of neural networks was compared with more established methodologies in the field of company failure, namely multiple discriminant analysis and logistic regression analysis.
A sample of twenty-eight failed companies and forty non-failed companies was selected and, for each of the five years prior to failure, seventeen financial ratios computed, Whilst the predictions of both neural networks and logistic regression appeared to be more accurate than those obtained by multiple discriminant analysis, it was shown that only the results comparing logistic regression with multiple discriminant analysis, five years before failure, were statistically different.
To investigate the appearance of trends in the data-set, neural networks were trained with financial ratios encompassing the five years prior to failure. However, no benefit was obtained in using ratios more than one year prior to failure.