Two-stage model for exchange rate forecasting by emd and random forest

Abstract

This study applied random forest (RF) and empirical mode decomposition (EMD) techniques to exchange rate forecasting. The aim of this study is to examine the feasibility of the proposed EMD-RF model in exchange rate forecasting. For this purpose, the original exchange rate series were first decomposed into a finite, and often small, number of intrinsic mode functions (IMFs) and one residual component. Then, a random forest model is constructed to forecast these IMFs and residual value individually, and then all these forecasted values are aggregated to produce the final forecasted value for exchange rates. The daily USD/NTD, USD/JPY, USD/HKD and USD/AUD exchange rates were employed as the data set. The experimental results are that MAPE for the four data sets are, respectively, 0.278%, 1.143%, 0.153% and 5.944%, which shows good performance according to the 10% threshold suggested by Lewis.