Financial Distress Prediction Based on Decision Tree Models

Predicting Corporation's financial distress accurately and efficiently is very important for banks, investors, enterprises and regulatory authorities. This paper analyzes the use of decision tree for corporate financial distress prediction. Linear models, although simple and easy to interpret, require statistical assumptions which may be unrealistic; meanwhile, neural networks are usually too complicated to comprehend. Decision tree, as one of the most efficient data mining methods, is not only able to discriminate patterns which are not linearly separable, but also can be easily understood. In this paper, an algorithm is proposed to select dilation and translation parameters that yield a decision tree classifier with good parsimony characteristics. The models are built in a case study involving both failed and continuing Chinese listed firms in the period of 2003-2005. The results, supported by a test study, show that decision trees may be a valid model to predict listed firms' financial distress in China.