The Application Of Machine Learning Method In Exchange Rate Forecasting

This paper selects a series of machine learning algorithms to figure out if macroeconomic variable effects in exchange rate forecasting.We apply several machine learning algorithms to predict the variation of exchange rate,and then construct a hedge strategy of a global assets portfolio.We set a random walk model and an OLS model as the benchmark of the accuracy of exchange rate forecasting,and a no hedge strategy and a full hedge strategy and an OLS model as the benchmark of the profit of our global assets portfolio investment.The results show that we get better accuracy in predicting exchange rate,and better risk-return ratio in global assets portfolio investment with some machine learning algorithms.More,we found that LASSO regression,random forest model and partial least squares algorithm are the best three algorithms in exchange rate forecasting.And a linear regression model based on gradient descent could help investors get more return rate and less risk with random forest model and partial least squares algorithm.