به كارگيري شبكه هاي عصبي مصنوعي در پيش بيني متغيرهاي اقتصادي و مقايسه آن با روش هاي اقتصادسنجي : پيش بيني روند نرخ ارز در ايران

A COMPARISON OF ARTIFICIAL NEURAL NETWORKS (ANN) WITH ECONOMETRIC METHODS FOR FORECASTING ECONOMIC VARIABLES: AN APPLICATION TO THE IRANʹS EXCHANGE RATE

The importance of forecasting trends in variables can be seen, obviously, in the policy making of economic rela­tions and in conducting appropriate activities. Accord­ingly, various methods have been developed to meet, the necessity of efficient prediction instruments. In the recent literature of applied economic studies, re­searchers have used widely the artificial neural network (ANN), particularly for prediction of monetary and fi­nancial variable trends. In this paper, we use the time series of Iranʹs exchange rate over the period 1959-2002, and employ two different methods of prediction in order to examine the hypothesis in which ANN is more efficient, and reliable than ordi­nary economic techniqnes in forecasting trend of Iranʹs exchange rate. The results obtained approve the fact that, if artificial neural network methods are specified correctly, they have better performance in forecasting exchange rates rather than that of other econometric methods.