Forecasting EUR/USD exchange rate with nonlinear autoregressive with exogenous input neural networks

Abstract

Forecasting currency exchange rates is a challenging task for Forex specialists and traders. This article examines projecting the EUR/USD exchange rate using data collected at 5-minute and 1-minute intervals. Any nonlinear function can be approximated using neural networks. The purpose of this study is to examine the utilization of a Nonlinear Autoregressive with Exogenous Input (NARX) neural network using the Matlab Neural Networks toolbox. Two models are constructed in order to determine the most effective model for forecasting foreign currency exchange rates. Both models construct and assess numerous NARX networks with varying designs in order to determine the optimal network structure. The research indicates that the majority of errors in Model 1 are between 2.6 and 3.4 pips. However, the majority of errors in Model 2 are between 2.2 and 2.4 pips, indicating that Model 2 makes more accurate predictions than Model 1.