Title: Comparison of several combined methods for forecasting Tehran stock exchange index

Authors: Ali Raoofi; Amir Hossein Montazer-Hojjat; Pouyan Kiani

Addresses: Department of Economics, Allameh Tabataba'i University, Tehran, Iran ' Department of Economics and Social Sciences, Shahid Chamran University of Ahvaz, Ahvaz, Iran ' Faculty of Economics, Management and Business, Tabriz University, Tabriz, Iran

Abstract: Forecasting economic and financial variables is of high interest to economic policy-makers in all countries. In this paper, the Tehran Stock Exchange Price Index (TEPIX) is estimated and forecasted using daily data for the period 22 May 2011 to 11 August 2011. To achieve that goal, various forecasting methods will be applied, including ARIMA, FARIMA, ANN and ANFIS models. Comparing the forecast accuracy of the models mentioned above, using forecast accuracy measures such as RMSE, MAE, MAPE and U-Thiel implied that the combined models of ANFIS and FARIMA have outperformed other models of forecasting daily stock indices. However, statistical comparison of forecast accuracy of different models using statistics such as Harvey, Leybourne and Newbold shows no significant difference between the forecast accuracy of these models.

Keywords: stock markets; stock index forecasting; ANNs; artificial neural networks; ANFIS; adaptive neuro-fuzzy inference systems; fuzzy logic; FARIMA; Iran; daily stock indices; forecasting accuracy.

DOI: 10.1504/IJBFMI.2016.080128

International Journal of Business Forecasting and Marketing Intelligence, 2016 Vol.2 No.4, pp.315 - 333

Received: 02 Mar 2016
Accepted: 15 Jul 2016

Published online: 03 Nov 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article