Modeling And Forecasting Exchange-Rate Shocks

This paper considers the extent to which the application of neural networks methodology can be used in order to forecast exchange-rate shocks. Four major foreign currency exchange rates against the Greek Drachma as well as the overnight interest rate in the Greek market are employed in an attempt to predict the extent to which the local currency may be suffering an attack. The forecasting is extended to the estimation of future exchange rates and interest rates. The MLP proved to be highly successful in predicting the shocks, while exhange-rates and interest-rates forecasts with MLP and RBF optimized by a genetic algorithm resulted in good approximations.