Research And Implementation Of Mobile Trajectory Prediction By Combining Complex Features

With the rapid development of broadband wireless access technology and mobile terminal technology, there is an urgent hope to be able to enjoy the Internet information and services in the process of moving anytime and anywhere, mobile Internet arises at the historic moment and rapid development. A variety of location-based equipment services are required to provide accurate location of a mobile device. At present, the reliability and the accuracy of the positioning system has been improved, but sometimes position system is difficult to accurately track the object due to the limitations of the GPS system, mobile device and wireless Internet. So we need a reliable method to predict the future positions of the mobile objects.In this paper we propose two trajectory prediction approaches for mobile objects including a trajectory prediction approach for mobile objects by combining semantic feature(SG) and an approach to probabilistic path prediction in dynamic environment(P3D). In the SG, geographic trajectories of all users are transformed semantic behavior trajectories firstly. Based on that, semantic trajectory pattern sets are mined. Common behavior of mobile users is analyzed in semantic trajectories and users are clustered based on the semantic behaviors similarity, by which geographic trajectory pattern sets are mined. Based on both mined semantic trajectory pattern sets of individual users and geographic trajectory pattern sets of similar users, STP- Tree and SLP- Tree are constructed. By indexing and partly matching on the two pattern trees, introducing a kind of weigh function, we can predict a result for a user’s recent mobile trajectories. P3 D has several advantages. First, the target trajectories to be predicted are known before the models are built, which allows us to construct models that are deemed relevant to the target trajectories. Second, unlike the lazy learning approaches, we use sophisticated learning algorithms to derive accurate prediction models with acceptable delay based on a small number of selected reference trajectories. Finally, P3 D can be continuously self-correcting since we can dynamically re-construct new models if the predicted movements do not match the actual ones.Experimental results on a large number of real and synthetic data sets show that the prediction and recall of SG methods are significantly improved compared to the state-of-the-art methods. P3 D can dynamically adjust results by self- tuning prediction continuously. Compared with no self- tuning prediction methods, the performance has improved obviously.