Research On Frequent Items Mining Technology Based On Trajectory Data

With the continuous development and popularization of sensor network(WSN), Global Positioning System(Global Positioning System, GPS) and wireless communication technology, a large number of the trajectory spatio-temporal data of mobile users has been accumulated in the application server. Based on the analysis of the users’ history trajectory data, many users can be found frequently in fixed time interval always appears within a fixed geographical scope. Mining the implicit and effective frequent patterns from the historical trajectory of mobile users is an important research topic in the area of data mining, which is important for some areas such as mobile data compression, analysis and forecast of mobile objects, personalized location-based services, the regional social behaviors mining, traffic management decision-making and business promotion and so on. Visiting time is ignored in the existing researches on the hot area found for mobile trajectory. To address the problem, the frequent pattern mining of trajectory data combined the attributes of space and time is studied.First, the existing region identification methods and existing problems are analyzed and an improved adaptive multi-granularity region identification method is presented. The trajectories of the mobile user is first segmented, and then walking and non-walking sections are distinguished, finally Clustering II algorithm can be used to cluster walking roads which can find different granularities space. The identified stay region will be the basis of the next step of frequent spatio-temporal items mining.Second, the concepts of spatio-temporal item and frequent spatio-temporal item combined the attributes of time and space are presented. Frequent spatio-temporal items can express that the user visit frequently which geographical areas in what time interval. A mining algorithm for frequent spatio-temporal items based on 3D grid is proposed. The three steps are used in the proposed method. In the first place, the spatio-temporal items can be generated. A clustering algorithm based on density is used to identify the stay regions of user and the elements of the spatio-temporal item that include User_id, Stay_ region, Time_ interval are expressed. There is one more point, I should touch on, that the spatio-temporal items are mapped to a 3D grid. Spatio-temporal items which contain the longitude, latitude, time will be mapped to predetermined 3D grids which integrate dimensions of time and space, and every support of the unit 3D grid can be recorded. The last but not the least, the 3D grid cells which meet the minimum support are extracted and the adjacent cells are merged to generate frequent spatio-temporal items.Finally, the experimental results on real trajectory data set show that the proposed approach can mine frequency spatio-temporal item with different supports from trajectory data. The mined frequent spatio-temporal items can reflect correctly the users frequently visit which geographical area in what time interval. The proposed approach has favorable performance.