Improving Business Failure Predication Using Rough Sets with Non-financial Variables

Abstract

Rough set models with financial variables have proven to be effective in predicting business failure. To enhance the predictive performance of rough set models, this paper includes a non-financial variable, auditor switching, into the modeling process, in addition to 14 financial ratios commonly used in business failure research. An empirical study on 62 failed firms and 62 one-to-one matching non-failed firms in Taiwan between 1998 and 2005 is conducted, using available data for the three years before failure. Six rough set models are constructed individually with and without the auditor switching variable, using the three-year data respectively. The empirical study shows that the non-financial variable is the most significant attribute and plays an essential role in enhancing the performance of rough set models. These findings highlight the effectiveness of rough set models for business failure prediction and particularly the importance of incorporating non-financial variables in business failure research.