Artificial neural networks have received more and more attention for exchange rate forecasting in recent years. However, the vast majority of research is done for freely floating currencies. In this paper neural network models are used to forecast the exchange rate of the Armenian dram, a non-freely floating currency. A number of different neural network architectures are evaluated. Back-propagation, modular neural networks, and general regression neural networks provide highly accurate predictions. The potential importance of results for relevant business communities as well as some of their shortcomings are discussed.
展开▼
