EFFECTIVENESS OF NEURAL-NETWORK TYPES FOR PREDICTION OF BUSINESS FAILURE

The study examines the effectiveness of different neural networks in p redicting bankruptcy filing. Two approaches for training neural networ ks, Back-Propagation and Optimal Estimation Theory, are considered. Wi thin the back-propagation training method, four different models (Back -Propagation, Functional Link Back-Propagation With Sines, Pruned Bock -Propagarion, and Cumulative Predictive Back-Propagation) are tested. The neural networks are compared against traditional bankruptcy predic tion techniques such as discriminant analysis, legit, and probit. The results show that the level of Type ! and Type II errors varies greatl y across techniques. The Optimal Estimation Theory neural network has the lowest level of Type I error and the highest level of Type II erro r while the traditional statistical techniques have the reverse relati onship (i.e., high Type I error and low Type II error). The back-propa gation neural networks have intermediate levels of Type I and Type II error. We demonstrate that the performance of the neural networks test ed is sensitive to the choice of variables selected and that the netwo rks cannot be relied upon to ''sift through'' variables and focus on t he most important variables (network performance based on the combined set of Ohlson and Altman data was frequently worse than their perform ance with one of the subsets). It is also important to note that the r esults are quite sensitive to sampling error. The significant variatio ns across replications for some of the models indicate the sensitivity of the models to variations in the data.