Financial Distress Prediction Using Adaptive Neuro-Fuzzy Inference System

This study describes a methodology for modelling corporate bankruptcy prediction by utilizing the application of the Adaptive Neuro-Fuzzy Inference System (ANFIS). For modelling in ANFIS, Firstly, input variables ranking by importance based on Neural Network algorithm and results using in corporate bankruptcy prediction modelling by ANFIS.Statistical population of this study is firms accepted in Tehran Stock Exchange during the time period between 2009 and 2014. Among the listed companies, companies that have the financial distress (in 2014) were selected according to the three criteria. The companyʹs stock symbol is closed in 2014; orDebt Ratio is close to one; orCompany in financial distress period (time period between 2009 and 2014) has net loss. After review, twenty financial ratios select as input variables of prediction model and calculated based on the financial statements of 41 selected companies in the time period between 2009 and 2013. Moreover, Debt Ratio (output variable) is calculated based on the financial statements of 41 selected companies in 2014.Artificial Neural Network selects firm size, stockholders equity to debts, earned capital, working capital to assets, working capital to sale, assets earning power, and operational cash to sale ratios as the important predictors of company financial distress with suitable estimate accuracy (88%). In the following, these variables were used as input variables to the ANFIS model.The training and check data sets are employed in the ANFIS model. ANFIS is trained with the help of MATLAB version R2009a. In this study, the back propagation method and Gaussian curve membership function were used for ANFIS modelling. The experimental result shows that the RMSE values for checking and training data are 0.1535 and 0.1539, respectively. The extracted ANFIS model integrates both neural network nonlinear capabilities and fuzzy logic qualitative approach and has potential to capture the benefits of both in a single framework. Nonlinear relationships between two financial ratios and output variable can be the basis for prediction of any corporate bankruptcy in future.