EPS Modeling and Prediction of Listed Companies in Tehran Stock Exchange with GMDH Neural Network Approach

Abstract

Predicting profits of per share and economic changes as long have been a favorite event for investors, managers, financial analysts and creditors. In this paper GMDH neural network approach is used as a tool with top capability in navigation and detection of complex nonlinear process with limited number of observations for modeling and prediction of profit of per share of listed companies in Tehran Stock Exchange. First a model was developed then by using deductive process and the exclusion of each variable from basic patterna total of eight models were run. The results showed patterns of exclusion return on assets, current ratio and efficiency of basic pattern of investment have the most effect in reducing prediction error respectively. Inventory turnover and collection period has also double effect in reducing errors. Finally the excellence accuracy of GMDH neural network in prediction of profit of per share compared to ARIMA method based on error criterion was approved.

Keywords

• Financial ratios
• Profits of Per Share
• GMDH Neural Network
• ARIMA Method