Evaluation of the reasons for business failures has been a major preoccupation of researchers and practitioners for many years. A large number of methods including multiple regression analysis, discriminant analysis, is, logit analysis, recursive partitioning algorithm, and several other techniques have been used for the prediction of business failures. This study has followed a different approach using the Rough Set theory to investigate the reasons of bankruptcies in the Turkish banking system. For this purpose, financial ratios of 29-41 commercial banks were examined for the 1995-2007 period. The results showed that this model is a promising alternative to the conventional methods for bankruptcy prediction.