The Usefulness of Random Forest Classifier and Relief Features Selection in Financial Distress Prediction: Empirical Evidence of Companies Listed on Tehran Stock Exchange

The Purpose of this research is investigating the usefulness of random forest classifier and relief features selection in financial distress prediction of companies listed on Tehran Stock Exchange. In this regard, through reviewing literature, 69 predictive features (variables) were specified as the initial features based on the popularity in the literature and the availability of the necessary data. By using relief method, optimal variables were selected from initial variables. In overall, the experimental results of investigating 95 financially distressed and 95 non-financial distressed in 2002 to 2014, indicated that random forest outperforms the logistic regression. In other words, the application of this classifier, increases the mean of accuracy, and reduces the occurrence of type I and type II errors. Furthermore, the results confirmed the usefulness of relief method in predicting financial distress. . In other words, using selected variables of this feature selection method (relative to using 69 initial variables) increases the mean of accuracy, and reduces the occurrence of type I and type II errors