Recent research has used financial ratios to establish the diagnosis models for business crises. This research explores a broader coverage of financial features, namely the recommended financial ratios from TEJ (Taiwan Economic Journal) database in addition to those financial ratios studied in prior literature. The aim of this research is to discover potentially useful but previously unaware financial features for better prediction accuracy. In this study, we had applied data mining techniques to identify five useful financial ratios, which two of them, tax rates and continuous four quarterly EPS are previously unaware to the research community. Our empirical experiment indicates that our proposed feature set outperforms those models proposed by prior scholars in terms of the prediction accuracy. (C) 2011 Elsevier Ltd. All rights reserved.