Financial failure prediction using data mining: An application in Istanbul Stock Exchange

Financial failure prediction is of great significance in terms of the long lasting operation and forecasting of the difficulties for the companies. Especially after 1970s, companies have started to experience more financial problems together with the increase in the risk concept paralleling the developments in the international financial system. In this context studies aiming to predict financial failure have become increasingly significant. The fact that theoretical findings in prediction studies could be put into practice in real life has increased the significance of the issue. The aim of the current study is to determine the factors concerning the prediction of the financial failure of the companies functioning in the manufacturing sector in ISE. To develop prediction models, recently popular data mining methods of decision trees C5.0 and neural networks techniques were used. The data of the study was composed of the 12-month balance sheets and income statements of the companies between 2005 and 2010. According to the findings, it was found that the most influential variable on financial failure is ?Operating Income / Total Assets? ratio. Moreover, the neural networks technique is better performing in financial failure prediction than the decision trees C5.0 technique.