Abstract: A location prediction algorithm based on movement tendency (LP-MT) was proposed, which not only buils moving object’s historical activity model borrowing from Markov thinking, but also added movement tendency as an important reference of location prediction. Besides, each of the historical stay regions was considered as a candidate of future location. According to the characteristics of the future location, algorithm results were divided into prediction locations and recommendation locations. Experiment results on real dataset show LP-MT method can ensure good efficiency and improve the accuracy by 10% compared with the existing algorithms.