An investigation on methods of artificial neural networks and decision tree estimation of financial failure

Financial failure may adversely influence stakeholders as well as overall economy. Therefore, financial failure prediction has become one of the most important fields of financial research. During years models have been developed using different methods to predict financial failure. The most popular and widely used methods are multivariate statistical methods and artificial neural networks. The purpose of this study is to predict financial failure of firms whose stocks were traded in Istanbul Stock Exchange (ISE) one year earlier by means of Data mining method. The prediction performance of these methods are compared and the best method is determined. Based on the data of firms whose stocks were traded in BIST, prediction methods were developed by means of data mining, decision trees, artificial neural networks, discriminant analysis and logistic regression. Then financial failures were predicted one year earlier. According to prediction performance, C&R tree, CHAID tree, logistic regression, artificial neural network, QUEST tree and discriminant model predicted 100%, 98%, 96.36%, 92.73%, 92.73% and 80% respectively. Among models C&R tree and discriminant model made the highest and lowest prediction percentage respectively. Keywords: Financial Failure, Data Mining, Artificial Neural Networks, Decision Tree