Abstract: In order to predict the uncertain trajectories in an efficient and accurate fashion, this paper introduces an uncertain trajectory prediction algorithm based on trajectory continuous time Bayesian networks (CTBN). It contains three essential phases: mining hotspot regions by partitioning trajectories into distinct hotspot clusters; constructing trajectory CTBN which is a states combination of three important variables including street identifier, moving speed, and moving direction; predicting the motion behavior of moving objects in order to obtain possible trajectories. Experimental results demonstrate that the proposed method can accurately predict the possible motion curves of moving objects in different trajectory data sets when compared with the naive prediction algorithm. In addition, experiments verify the essential role of hotspot region mining, which can help save prediction time at about 60% with a guarantee of high prediction accuracy.