Comparing the performance Of Artificial Neural Networks(ANN) and Auto Regressive Moving Average(ARIMA) Model in Modeling and Forecasting Short-term Exchange Rate Trend in Iran

Exchange rate and its related fluctuation plays a significant role as one of the most important issues of each country's foreign trade sector. Many factors such as economic, politics, and psychological factors impress on exchange rates and these factors create more uncertainty situations. Policymakers’ attempt is to reduce this uncertainty via forecasting this variable with minimal error.Artificial neural networks have high potential in modeling complex processes and forecasting dynamic nonlinear paths .Therefore, in this study has tried to use the artificial neural network (ANN) In addition to modeling and forecasting daily exchange rates during the period of March 2002 to March 2005, and minimizing the forecast error by this method, its results were compared with that of ARIMA based on forecasting accuracy evaluation criteria , and to examine the sensitivity of model results toward exchange rates.Estimation of the model with the same method for three data sets exchange rate including dollar,euro and pound have been performed .Results indicate that used neural network has better predictive power in comparison with arima model while pound and Euro exchange rates’ prices are function of their yesterday prices and dollar exchange rate price is a function of its price over the past 6 days .