Long term prediction of Tehran price index (TEPIX) using neural networks

It has been previously shown that the dynamics governing the share prices in Tehran Stock Exchange can be considered as a chaotic time series. Due to the initial sensitivity of the price generating process, it is shown that linear classical models such as ARIMA and ARCH are not able to efficiently model the dynamic of share prices in Tehran stock exchange for long term prediction purposes. However, non-linear neural network models are proposed to model the Tehran price index (TEPIX) daily data process and it is shown that such nonlinear models can successfully be used for the long term prediction of TEPIX daily data. Real data for the period of 1996 to 1999 are used to validate the prediction results.