TABLE OF CONTENTS Throughout the last decade, the application of Artificial Neural Networks in the areas of financial and economic time series forecasting has been rapidly expanding. The present chapter investigates the ability of Distributed Time Lagged Feedforward Networks (DTLFN), trained through a popular Differential Evolution (DE) algorithm, to forecast the short term behavior of the daily exchange rate of the Euro against the US Dollar. Performance is contrasted with that of focused time lagged feedforward networks, as well as with DTLFNs trained through alternative algorithms.
Keywords: Artificial neural networks, differential evolution algorithms, time series prediction.
A central problem of science is forecasting; how can knowledge of the past behavior of a system be exploited in order to determine its future evolution. As of 1997 the foreign exchange market constitutes the world's largest market with daily transactions surpassing 1 trillion US dollars on busy days. More than 95 percent of this volume is characterized as speculative trading, i.e. transactions performed in order to profit from the short term movement of the exchange rate.
Two schools of thought compete in the field of financial forecasting, fundamentalists and technical...
