Research On Trajectory Community Detection Based On Spatial-Temporal Similarity And Semantic Similarity

The popularity of mobile portable equipment and the rapid development of wireless communication technology and GPS technology make it possible to track moving objects.With the rapid growth of trajectory data of moving objects,more and more researchers do a lot of contribution on trajectory data analysis and mining.Trajectory clustering is one of the hot points in trajectory mining.At present,few trajectories clustering algorithm uses the multiple features of a trajectory and the community detection algorithm for graph data.Trajectory clustering with multiple features of a trajectory can make the clustering result more accurate.And the trajectory clustering based on community detection algorithm can save more computing time and detect the number of communities automatically.Therefore,the thesis proposes the trajectory community detection method based on spatial-temporal similarity and semantic similarity.In order to achieve highly accurate trajectory similarity,the thesis proposes two methods to measure the spatial-temporal similarity and semantic similarity of trajectories.Both of the measure methods consider the time feature to achieve more accurate similarity.The main contributions of this thesis are summarized as follows:Firstly,the thesis proposes the semantic trajectory histogram method to measure the trajectory semantic similarity,which extracts effective semantic feature and takes the time feature into consideration.In addition,the notion of semantic trajectory transition frequency is proposed.By comparing the transition frequency of different semantic trajectories,the similarity measure result can be more accurate.Secondly,the thesis proposes the weighted longest common subsequence method to measure the trajectory spatial-temporal similarity.Compared with the traditional longest common subsequence method,it adds the time-based weight to consider the effect of trajectory time feature.In addition,the notion of minimum bounding grid rectangle is proposed to accelerate the calculation of the spatial-temporal similarity.Thirdly,the skyline detection method is used to make trajectory features clear and the accuracy of the clustering result is improved.Finally,the community detection algorithm for the weighted undirected graph is used for trajectory clustering to improve the clustering result.The theoretical analysis and experimental evaluations show that the trajectory community detection method,which is based on spatial-temporal similarity and semantic similarity,is feasible and accurate.