Reserch On Prediction Technique Of Moving Object Location

Position prediction has become an extremely important research topic in the location management of moving objects. The prediction methods have been proposed and improved continuously. In recent years, an approach based on frequent trajectories is introduced, which achieves its goals through mining history trajectories data of moving objects and matching its current moving trends.The novel introduced technique evolutes from association rules of data mining, which still needs improvements. The main work and contributions of this dissertation are summarized as follows:Firstly, a novel algorithm TidTraj based on trajectory id list is proposed. Since the existing techniques scan database excessively. However, algorithm TidTraj makes up the shortcomings of the traditional methods. It scans the database only once, which obviously improves the efficiency of prediction on processing time. Extensive experiments results on traffic trajectories dataset demonstrate the efficiency and effectiveness of the proposed algorithm for prediction.Secondly, a new algorithm for frequent trajectories mining TidTrajUP is provided based on the incremental update of the database to predict, to settle the problem that the AprioriTraj mechanism does not take the corresponding strategy into account when the database is incrementally updated. TidTrajUP only mines part of the frequent trajectories, so it makes full use of the original frequent trajectories datasets, avoids redundant operations and improves the efficiency of mining. The comparative experiments results dedicate the efficiency of the algorithm.Thirdly, a concept of time efficiency factor is introduced in this paper to improve the accuracy of the predicting, considering the impact of time on the objects'moving rules as time going on. Time efficiency factor is added to the computation of support for frequent trajectories as a description for the effect of time on movement. By adding this parameter, support can fully reflect the influence of time on the accuracy of prediction. Focusing on the support of time efficiency factor and trajectory id list, another frequent trajectories mining algorithm Agtraj is proposed based on time efficiency factor as a kernel algorithm for the first phrase prediction and also a complete prediction process is given in this paper. Certain improvement on the prediction accuracy is made compared with the traditional methods. Experiment results show the improvement of the accuracy.