Dynamics of Modeling in Data Mining: Interpretative Approach to Bankruptcy Prediction

This paper uses a data mining approach to develop bankruptcy prediction models suitable for normal and crisis economic conditions. It observes the dynamics of model change from normal to crisis conditions and provides interpretation of bankruptcy classifications. The bankruptcy prediction model revealed the major variables in predi cting bankruptcy to be ‘cash flow to total assets’ and ‘productivity of capital’ under normal conditions while ‘cash flow to liabilities’, ‘productivity of capital’, and ‘fixed assets to stockholders equity and long -term liabilities’ under crisis conditions. The accuracy rates of final prediction models in normal conditions and in crisis conditions were found to be 83.3%, and 81.0%, respectively. When the normal model was applied in crisis situations, prediction accuracy dropped significantly in the case of bankruptcy classification (from 66.7% to 36.7%) at the level of a blind guess (35.71%). Therefore, the need for a different model in crisis economic conditions is justified.