E³TP: A Novel Trajectory Prediction Algorithm in Moving Objects Databases

Abstract

Prediction of uncertain trajectories in moving objects databases has recently become a new paradigm for tracking wireless and mobile devices in an accurate and efficient manner, and is critical in law enforcement applications such as criminal tracking analysis. However, existing approaches for prediction in spatio-temporal databases focus on either mining frequent sequential patterns at a certain geographical position, or constructing kinematical models to approximate real-world routes. The former overlooks the fact that movement patterns of objects are most likely to be local, and constrained in some certain region, while the later fails to take into consideration some important factors, e.g., population distribution, and the structure of traffic networks. To cope with those problems, we propose a general trajectory prediction algorithm called E³TP (an Effective, Efficient, and Easy Trajectory Prediction algorithm), which contains four main phases: (i) mining “hotspot” regions from moving objects databases; (ii) discovering frequent sequential routes in hotspot areas; (iii) computing the speed of a variety of moving objects; and (iv) predicting the dynamic motion behaviors of objects. Experimental results demonstrate that E³TP is an efficient and effective algorithm for trajectory prediction, and the prediction accuracy is about 30% higher than the naive approach. In addition, it is easy-to-use in real-world scenarios.

This work is supported by the National Natural Science Foundation of China under Grant No. 60773169, the 11th Five Years Key Programs for Science and Technology Development of China under Grant No. 2006BAI05A01, the Youth Software Innovation Project of Sichuan Province under Grant No. 2007AA0032 and 2007AA0028.