FORECASTING OF EARNINGS PER SHARE FOR ACCEPTED FIRMS IN TEHRAN'S STOCK EXCHANGE BY UTILIZING THE GENETIC ALGORITHM OF ARTIFICIAL NEURAL NETWORK

Abstract

Forecasting the Earnings per Share for Investments is Particularly Important because it is considered an Important Factor in Share assessment methods and a Fundamental Factor in making Investment decisions. In Order to Forecast earnings per share using an Artificial Neural Network, 61 Firms were selected in eight Financial Years From the Beginning of 2000 to the End of 2007 along With 9 Variables (8 Input Variables and 1 Output Variable), yielding 4392 (9 x 8 x 61) data points. The Research Hypothesis is that a Neural Network with a genetic algorithm can forecast earnings per share. In order to Test the Hypothesis, MATLAB Software was used to determine the Mean Square Error and Mean Absolute Error. The researchers' hypothesis is supported.