Forecasting the euro/dollar changes using predicted values of stock index: selected econometric models and multilayer perceptron

In the paper, we discuss the results of euro-dollar exchange rate prediction. Forecasts are made on the basis of the dynamic econometric model and multilayer perceptron. Forecasting is provided in a two-stage procedure. In the first step Dow Jones STOXX 50 Index is predicted applying ARMA model constructed for the rates of return of this variable. In the second stage the predicted values of Dow Jones STOXX 50 Index are introduced to the econometric and neural network models, that are constructed for the eurodollar rate of returns. Then the forecasts obtained are transformed into euro-dollar exchange rate. Data included in the analysis covers the period from 5th January 1999 to 13th July 2001. However, the estimation is made for the period till 29th June 2001 and the forecasts are made for the next ten sessions.