Predicting Tehran Securities Stock Index By Using Neural Networks

The size and process of the stock price indices are among the most important factors affecting the decisions of the investors in the financial markets. In order to predict the market, different techniques have been used, the most common of which are regression methods and ARIMA models. However, these models have been unsuccessful in the practical prediction of some series.
In the present research, in order to predict the total index of the stock, the Feed Forward Neural Network model with the law of back propagation was used in three networks with different input models, and the results of the model were compared to the result of multi – variable regression models and ARIMA models.
The results indicated that the neural network method showed considerably fewer RMSE errors than RMSE errors in other methods, and that in Tehran stock market short – term prediction within a shorter interval is more suitable than long – term prediction within a longer interval.