This study compared the in-sample and out-of-sample forecasting performance of the asset market approach model, the random walk model, and the ARIMA model for the Korean won/U.S. dollar exchange rate over both short-term and long-term horizons. The out-of-sample forecasting results showed that the random walk model exhibited the most superior forecasting power at the 1-month and 3-month horizons, while the real interest rate differential model and the composite model were found to be superior at the 6-month and 12-month horizons. The in-sample explanatory power results revealed that the random walk model provided the most accurate won/dollar exchange rate forecasts across all time periods. Therefore, it can be concluded that for short-term forecasting of the won/dollar exchange rate, the random walk model should be employed, while for long-term forecasting, the real interest rate differential model and the composite model can achieve greater forecasting accuracy.