This study evaluates the performance of international asset allocation strategies based on predictions of foreign exchange rates and stock market returns. We utilize various machine learning models and a wide range of economic and financial variables to predict the KRW - USD exchange rate and U.S. and Korean stock market returns. Our findings suggest that machine learning models outperform benchmark models in predicting both the exchange rate and stock market returns. Furthermore, a machine learning-driven global portfolio that accounts for exchange rate fluctuations demonstrates enhanced performance. This study presents empirical evidence substantiating the use of machine learning techniques to forecast foreign exchange rates and construct a compelling global portfolio.