Prediction of Stock Index Using Wavelet Neural Networks in Tehran Stock Exchange

In this research, the total equities in Tehran Stock Exchange are predicted using different neural network models. Research hypothesis states that the performance of WDBP wavelet neural network model for stock index prediction is better than PB neural network model and research period is ten years from the beginning of 2002 to the end of 2011. Information is collected from the statistics and data in Tehran Stock Exchange’s database. In order to create the WDBP model, db5 wavelet is used for denoising the data in five steps. The criterion used for measuring the prediction error is root mean square error (RMSE). Wilcoxon hypothesis test is conducted on results after prediction by neural networks. Test results indicate that the significance level of overall index is less than 0.05. Therefore, the null hypothesis is rejected. In other words, there is a significant difference between the prediction of neural network method and that of wavelet neural network.