Comparing the Exchange Rates Predicted by STAR Non-linear Models and Alternative Models

Exchange rate known as a strategic variable plays an important role in the economy, because of affecting on different sectors in economy all over the world. So, exchange-rate predictions have always been an important subject for the researchers in Economics. This paper tries to study the attributes of exchange rate developed by monthly official data of Iran Stock Exchange based on Smooth Transition Autoregressive (STAR) models. The result of simulation based on STAR models and estimated by Genetic Algorithm method, outperforms linear time series models, such as ARIMA out of sample predictions based on RMSE, MAE and DA criteria.