Exchange Rate Forecasting using Recurrent Self-Organizing Maps

Exchange rate is one of the most important variables in economics, which impacts many other macroeconomic variables in an economy.  Such changes are more important in developing countries, which are often exporters of raw materials. Exchange rate fluctuations in Iran is also critical since the country is highly dependent on foreign exchange earnings from oil exports and also subject to the severs sanctions imposed by the US in recent years. In this paper, we propose a forecasting model using recurrent self-organizing maps to forecast the exchange rates considering the factors affecting the exchange rate markets. In total, 12 selected variables are used in modeling the exchange rate, which includes the effective macro variables and the price index of competing markets, such as gold, stock exchange and housing. We also control for the effect of sanctions. The results show that the model produces relatively accurate forecasts of the exchange rates with a less than 3% error.