Composite Classifiers for Bankruptcy Prediction

Business failures can cause financial damages to investors, creditors, or even society. For this reason bankruptcy prediction is one of the most challenging tasks in the field of financial decisionmaking. Business failure prediction has been an active research area since the 60s. The work of Beaver (1966) who performed univariate analysis of financial ratios and the work of Altman (1968) who employed Multiple Discriminant Analysis (MDA) mark the starting point of the relevant research. In the following years many researchers have proposed statistical and artificial intelligence techniques to predict bankruptcies. In most cases the artificial intelligence techniques outperformed the statistical techniques. In the current research one can notice that there is a strong trend concerning the proposed classification methods. This trend is to design and apply composite classifiers, i.e. hybrid and ensemble classifiers. According to the results, composite classifiers perform better than the single classification techniques. The purpose of the present chapter is to cover the topic of the design and application of composite classifiers for bankruptcy prediction. Key issues related to composite classifiers are analyzed to familiarize the reader with these techniques. Subsequently nine selected papers that were recently published in high reputation journals of an impact factor score of more than 0.5, are presented. This presentation concentrates on problems, methodological issues, proposed designs and findings that refer to the employment of composite classifiers for bankruptcy prediction. Other important issues relevant to bankruptcy prediction, such as improved single classifiers, alternative input data, feature selection etc are beyond the scope of the present article.