Comparison of The Performance of Artificial Neural Network Models for Exchange Rate Prediction in Iran

Given the high exchange rate fluctuations in Iran, its prediction is one of the major issues and challenges for different groups in the country. This study investigates the performance of six different static and dynamic neural networks for forecasting exchange rates using fundamental, technical and hybrid approaches and using seasonal data over the period (1) 2004- (4) 1396 for variables influencing exchange rate including inflation, Liquidity and GDP for the two countries, Iran and the United States. The findings show that the number of neurons did not have a regular effect on the performance of the networks and that the best results occurred at breaks of three and four. The results also show that the best performance of the static neural network is achieved by a technical approach with a structure of sixteen neurons and four interruptions which provides a relatively accurate exchange rate prediction despite the low number of input data.