Abstract
In order to improve the exchange rate forecasting ability, we use the Bayesian Treed Gaussian Process model with different methods to obtain different trading strategies, and employ some kinds of economic variables as fundamentals for exchange rate forecasting. Therefore, the aspects of our research is divided into two parts. The first part is whether there is any difference in exchange rate forecasting ability when the trading strategies obtained by the different classification standards of countries. The second part is whether economic variables as fundamentals will help increase the exchange rate forecasting ability. Directional Accuracy、Excess Predictability、 Annual Percentage Rate and Sharpe Ratio are used to measure the exchange rate forecasting ability. The paper finds that different classification standards of countries does not affect the exchange rate forecasting ability; economic variables all help to improve the exchange rate forecasting ability. We also find that t he forecasting results of the Bayesian Treed Gaussian Process model dominate those of Random Walk and Random Walk with Drift. In particular, the exchange rate forecasting ability is the best when we employ trading strategies, stock prices and oil prices as fundamentals.