Application of Dynamic Parametric and Non Parametric Systems in Stock Market Return Forecasting: Case Study of Tehran Stock Market

Paying attention to the role of capital markets in providing suitable floor for capital gathering and distribution of financial resources, such markets and stock markets in particular are in the center of attention of domestic and foreign investors as well as government sector. One of the major issues in such markets is risk management that in the stock market this problem is closely related to the price or return forecasting and its importance has reflected in evaluation of market informational efficiency. in this regard, the present study focuses on two different (a)dynamic-parametric method of ARMA-PGARCH and (b): dynamic-nonparametric procedure of NARX artificial neural network and forecasting Tehran stock return. Predictions are carried out in the form of in-sample and out-sample using daily observations of TEPIX from Septemper-1997 to May-2015. Forecasting horizon of next five working days has adopted for the out-sample prediction and eight error criteria are picked out in order to assess accuracy of each approach. Outcomes of this research implying higher precision of the dynamic neural network performance in comparison with the parametric method of ARMA-PGARCH. In addition, the results are in favor of inexistence of weak-form of informational efficiency in Tehran stock market.