Anticipation of Iran Mercantile Exchange (IME) gold coin price using Artificial Neural Network Approach with GMDH Algorithm

Subject Areas : Applied Economics

عباس معمار نژاد ¹ , وحید فرمان آرا ²

1 - ندارد
2 - مسئول مکاتبات

Received: 2011-07-09 Accepted : 2011-10-25 Published : 2011-09-23

Keywords: Anticipation, Gold Coin Price, Macroeconomics Variables, Artificial Neural Network, GMDH Algorithm,

Abstract :

The economy of every country is composed of different sectors in which, the relationship amongst them determines the dimensions of the economy of that country. The capital market together with money market make up the financial market as the main arteries of an economy. Their operation has a significant influence on the growth and development of the economy. In cases where there is no constructive relationship between the financial market and economic sectors, economic performance might be subject to distortions. The stock market as a fundamental pillar of the financial market plays a crucial role in facilitating investments in the capital market. Given the importance of expectations in different economic fields, the main purpose of this study, as its title explains, is to anticipate of Iran Mercantile Exchange (IME) gold coin price Therefore, after a brief review of dominant economic theories, a new method, artificial neural network GMDH, is used to forecast the impact of macroeconomic variables( including the rate u.s. dollar as foreign exchange, the price of gold coin, the price of gold and oil in termes of dollar, the over-all index of stocks, the delivery date of gold coin) on the gold coin price. The GMDH Algorithm is a nonlinear model to anticipate complex systematic relationships between variables of the model. The special feature of this deductive algorithm is recognition and screening of the most effective variables to estimate the model with training samples and omit the non-significant ones from the simulation process with testing samples. So, an attempt is made to solve the model via iterative methods to minimize the typical standard Error like RMSE, MAPE, and so on.