The main goal of this thesis is to propose a model for forecasting Mobarakeh Steel Company productsprice in Iran Mercantile Exchange using artificial neural networks. Nowadays with increasing complexity of financial markets, market participants need efficient methodsfor prediction of future prices. Accurate price forecasting for steel commodities can have significant decision making implications for Mobarakeh Steel Company. This company is the biggest steel producer in Middle East and Northern Africa and one of Iran’s leading companies. In the first stage, we identified variables which effect on dependent variables by using expert opinions and previous researches. In the next stage effective variables and their relation with dependent variable have been analyzed by using self-organized maps. After considering forecasting methods and concerning unique specifications of artificial neural networks in forecasting such as their capability in modeling non-linear relationships between variables, we proposed an artificial neural network model for forecasting. We have designed a multi-layer perceptron neural network for forecasting of each steel product. The structure of the neural network is a three-layer back propagation (BP) network. The results show that proposed model provide less than 3 percent error in the test sets for all products. Accuracy of proposed model has been compared with multiple regression model and general regression neural networks (GRNN). Results indicate that the neural network model is more accurate than linear regression and GRNN.