Considering the multimodal nature of transport systems and cross-modal correlations, there is a growing trend of enhancing demand forecasting accuracy by optimization with multiple transport modes data jointly, which can improve accuracy but be less practical when different parts of multimodal datasets are owned by different institutions who cannot directly share dat. This study proposes an Unsupervised Knowledge Adaptation Demand Forecasting framework, which forecasts the demand of one mode (i.e., the target mode) by utilizing a pretrained model based on data of another mode, but does not require direct data sharing of another transport mode (i.e., the source mode). The unsupervised knowledge adaptation strategy is utilized to form the sharing features for forecasting by making the pre-trained network and the sharing extraction network analogous. Our findings illustrate that unsupervised knowledge adaptation by sharing pre-trained models to the target transport mode can improve the forecasting performance instead of direct data sharing.