Forecasting stock price using artificial neural networks: A multi-layer perception model - Iranian evidence

Abstract

In recent years, steel manufacturers have been playing a great role in economic growth, bringing about large amounts of stock exchange transactions in this industry. In the current study, we try to design a model to forecast stock price of steel industry, using artificial neural networks. To design the model, we used a three-layer network (five neurons in input layer, twelve neurons in the middle layer and one neuron in output layer), a sigmoid transfer function, 7% Alpha, 2% Etta and Windows Neural Network (WNN) software. The input variables of the network include net assets, P/E ratio, dividend per share (DPS), earning per share (EPS), amount of stock transactions, and stock price network output of companies being studied. The results from designed model show that if an artificial neural network is taught correctly, it can recognize the relationship between variables and it can help to forecast the stock price of steel industry with minimum error (35% in this research). Investors can forecast the stock price of steel manufacturing companies using these inputs variables and WNN software. © 2011 Academic Journals.