Financial distress prediction of Tehran Stock Exchange companies using support vector machine

Abstract

The main objective of this study is to evaluate and to compare the power to predict company financial distress by utilizing the Support Vector Machine (SVM) to the multiple-discriminant analysis and the logistic regression models. Companies approved for acceptance into Tehran Stock Exchange Market between 2007 and 2013 comprise the statistical population for the study. In order to predict financial distress based on financial ratios such as profitability, activity ratio, ratios per share, etc. by using the Support Vector Machine (SVM), the sample data has been divided into two separate groups: the training group and the experimental group. The training set is made up of 540 year-company and the experimental set is comprised of 120 companies in 2013.  Finally, conclusions obtained from SVM, multiple-discriminant analysis and the logistic regression models for predicting financial failure were surveyed and compared. Results of testing hypothesis indicate with a 95% certainty ratio that there is a significant difference in the average prediction accuracy of the three models. Consequently among the three, the SVM model has the highest accuracy level for predicting company financial failure and the multiple-discriminant analysis model has the lowest.