Forecasting foreign exchange markets: further evidence using machine learning models

Abstract

This study aims at examining the predictability of the autoregressive integrated moving average and deep learning methods consisting of the artificial neural network, recurrent neural network, long short-term memory (LSTM), and support vector machine. We will use these tools to estimate the parameters for predicting the accuracy of the foreign exchange returns. This study compares the forecasting performance between the autoregressive integrated moving average and deep learning methods. The comparison is based on the mean absolute percentage error, the root-mean-squared error, the mean absolute error, and Theil U. The empirical results indicate that the LSTM seems to outperform the other deep learning models as well as the traditional regression models.