The problem of personalized next point-of-interest (POI) recommendation is significant and of practical value in location-based social networks (LBSNs). Twitter together with other online social networks has begun to collect hundreds of millions of check-ins which capture the spatial and temporal information of user movements and interests. Due to the sparsity of data in regard to check-in, POI recommendations remain a challenging problem. An efficient point-of-interest recommendation system with Upstream Spatiotemporal Topic Model (USTTM) for LBSNs is proposed in this paper. This recommendation system contains purpose prediction phase that classifying the POIs in spatio-temporal database based on purpose and constructing a purpose ranking model to model the selection of user’s intended purpose for the trip. Second phase is scoring of each candidate POI which considers the properties of spatial and temporal information when calculating the score. USTTM is used in this system to model and analyze the spatio-temporal aspect of check-in data that can discover a user’s choice of region. Extensive experiments were conducted and the results demonstrate that the recommendation accuracy of the model outperforms the state-of-the-art POI recommendation models with good runtime performance. In quantitative analysis, effectiveness of USTTM in terms of accuracy of POI recommendation and accuracy of user and time prediction are evaluated and results show that the USTTM achieves better performance than the state-of-the-art models.