Mining Frequent Trajectory Based On Spatio-Temporal Data

With the rapid development of positioning and communication technology,a large amount of trajectory data has been collected.It gets more and more attention to excavate the hidden regularly movement and patterns of moving objects.To solve the problem that the semantic regularity and pattern is discovered insufficiently using single time granularity,we propose a spatio-temporal frequent trajectory pattern mining method based on multidimensional time granularity.The mainly content of this thesis is as follows:1.The premise of mining frequent regions and trajectory patterns is to recognize the stay regions in multidimensional time granularity.In this thesis,we propose a method for mining the spatio-temporal stay regions based on multidimensional time granularity(MTG_SR).Firstly,dividing the time granularity hierarchy.Then,preprocessing the trajectory data,to remove unrelated and redundant data.Finally,in order to recognize the stay regions in multidimensional time granularity,we adopt the strategy that combine sliding time window with the self-adaptive recognize saty regions method.2.In this thesis,we propose a spatio-temporal frequent region mining method with multidimensional time granularity(MTG_FR).First of all,in order to avoid the hard boundary problem,we propose a frequent regions recognizing method based on the idea of set theory.And then,we propose a self-adaptive time interval selection method based on the Gaussian mixture model,determining the time interval including many frequent spatio-temporal regions,so that we can solve the problem of using a sliding time window strategy is difficult to identify the time interval that people really focus on with more frequent regions.Finally,after visualizing the the spatio-temporal frequent regions,the moving objects' regular activity is discovered in multidimensional time granularity.The experimental results show that combing with using single time granularity,the method we proposed can fully extract the objects' regularly semantic movement adaptively.3.To solve the problem that the group semantic movement pattern information is discovered insufficiently,we propose a frequent trajectory pattern mining method based on multidimensional time granularity(MTG_FTP).We first propose two algorithms MTG_GSP and MTG_Prefix Span,based on classical sequence pattern mining method GSP and Prefix Span,to mine trajectory pattern.Then,employing the method self-adaptive time interval selection we proposed,Gaussian mixture model is used to select the time interval containing more frequent trajectory patterns.Finally,since visualizing the spatio-temporal frequent trajectory pattern,the semantic information of group movement and patterns is obtained.The experimental results show that combing with a single time granularity,the method we proposed can adaptively find out the group moving pattern with time comprehensively.