Efficiency compared to ARIMA and ARFIMA models for modeling and prediction of Tehran Price Index (TEPIX)

Document Type : Original Article

Authors
Habibollah salarzehi 1
Mansoor kasha kasha 2
Seyed-Hasan Hosseini 2
Mohammad Donyaei 3
1 Sistan & Baluchestan University, Zahedan, Iran
2 MSc in Financial Management, Sistan & Baluchestan University, Zahedan, Iran
3 Department of Management, Young Researchers Club, Zahedan Branch, Islamic Azad University
Abstract
This article examines the forecast performance of ARFIMA and ARIMA models using data on daily stock price index of Tehran in period 25/11/2001 to 30/11/2011. To estimate the d parameter and other parameters, the NLS method in the software package Oxmetric / pcgive was used. After comparing the results of research models, ARFIMA models based on AIC, the model was found superior in modeling TEPIX. Also we use naive methods for estimating the prediction. Comparing the accuracy of the prediction models by criteria such as MAPFE and RMSFE and confidence intervals of  the real values, we can deduce that the first Performance difference between the predicted long-term memory ARFIMA model is very minor compared to the ARIMA model And Secondly, inefficient ARFIMA model in Tehran capital market forecast is quite evident.
Keywords
Forecast
Return
Autoregressive Integrated Moving Average
Autoregressive fractionally integrated moving average