Developing a bankruptcy prediction model via rough sets theory

Abstract

The high individual and social costs encountered in corporate bankruptcies make this decision problem very important to parties such as auditors, management, government policy makers, and investors. Bankruptcy is a worldwide problem and the number of bankruptcies can be considered an index of the robustness of individual country economies. The costs associated with this problem have led to special disclosure responsibilities for both management and auditors. Bankruptcy prediction is a problematic issue for all parties associated with corporate reporting since the development of a cause–effect relationship between the many attributes that may cause or be related to bankruptcy and the actual occurrence of bankruptcy is difficult. An approach that has been proposed for dealing with this type of prediction problem is rough sets theory. Rough sets theory involves a calculus of partitions. A rough sets theory based model has the following advantages: (1) the rough sets data analysis process results in the information contained in a large number of cases being reduced to a model containing a generalized description of knowledge, (2) the model is a set of easily understandable decision rules which do not normally need interpretation, (3) each decision rule is supported by a set of real examples, (4) additional information like probabilities in statistics or grade of membership in fuzzy set theory is not required. In keeping with the philosophy of building on prior research, variables identified in prior recursive partitioning research were used to develop a rough sets bankruptcy prediction model. The model was 93% accurate in predicting bankruptcy on a 100-company developmental sample and 88% accurate on the overall separate 100-company holdout sample. This was superior to the original recursive partitioning model which was only 65% accurate on the same data set. The current research findings are also compared, both in terms of predictive results and variables identified, to three prior rough sets empirical bankruptcy prediction studies. The model produced by the current research had a significantly higher prediction accuracy on its validation sample and employed fewer variables. This research significantly extends prior rough sets bankruptcy prediction research by using a larger sample size and data from U.S. public companies. Implications for both bankruptcy prediction and future research are explored. Copyright © 2000 John Wiley & Sons, Ltd.