Earnings Per Share Forecast Using Extracted Rules from Trained Neural Network by Genetic Algorithm

Earnings per share (EPS) is one of the main financial ratio that is considering by managers, investors and financial analysts. It is usually using in investment decisions, profitability evaluation, profit risk, and stock price estimation. Therefore, EPS forecasting is a valuable and attractive task for managers and investors. This paper examines EPS forecasting using multi-layer perceptron (MLP) neural network and rule extraction from neural network by genetic algorithm technique and determined an optimal model between MLP and RE technique by evaluating their forecasting accuracy. For this purpose, we use 990 listed firms in Tehran Stock Exchange in the period of 2000–2010. The results show that the RE technique is significantly more accurate than the MLP model.