In order to plan high-speed rail transport services efficiently, it is necessary to be able to forecast fluctuations in passenger demand based on historical ridership data. Forecasting is difficult however, because of the number of components making up passenger demand. An effective way to forecast demand therefore should be to decompose these fluctuations into several independent demand components, which can then be forecast individually. This study applied an independent component analysis to decompose the fluctuation into several independent components. A method was then developed to forecast the fluctuation in passenger demand based on actual ridership data, calendar array, and number of people mobilized for large events.