Modelowanie kursu euro/dolar: dynamiczne modele ekonometryczne i sztuczne sieci neuronowe

Abstract

In the paper we present the results of the research regarding relationship among euro/dollar exchange rate and other variables from financial and good markets. The analyzed time series are observed daily. In the first stage we investigate selected variables from financial and good markets that influence the euro/dollar exchange rate applying the Granger causality test. Our investigation shows that futures contract influences the euro/dollar exchange rate the most. In the second stage we construct the econometric models (ARMA-DL and ARMADL with GARCH) and artificial neural networks (Multilayer Perceptron) that are used to modelling and forecasting of the euro/dollar exchange rate. In the last stage we compare different models using goodness-of-fi t measures, Theil inequality statistic and its decomposition