Comparing the precision of approaches of support vector machine and artificial neural networks to predict the benefits per share of listed companies in Tehran Stock Exchange

Stockholders for making proper investment decisions need information which assists them in making the best decision. Among available information، Information related to the per share forecasted earnings is the important ones in the user’s opinions. Beside، companies try to forecast the earnings per share with maximum accuracy to attract investors. Accordingly، the present study seeks to provide a model to improve earnings per share forecast of companies listed in Tehran Stock Exchange (TSE) using modern artificial neural networks approaches. For this purpose، first the factors affecting future earnings per share were inferred from internal and external research، then using the sample companies'' financial information in the years 2005-2012 and employing support vector machines and artificial neural networks methods، earnings per share forecasting was designed. Support vector machines model was able to forecast the sample companies’ next year’s earnings per share with an adequate error of 5%. This model introduces current year’s earnings per share with effective coefficient of 25% as most effective variable to forecast next year’s earnings per share. The results suggested that the support vector machines model has similar performance in comparison with artificial neural networks model.