The Research On Social Network Spatiotemporal Periodic Behavior Pattern Mining Algorithm

With the rapid development of the Computer, Internet and CommunicationTechnology, social activities and the people linked to others are more closely andfrequently. As the network structure has become more and more complicated, thedeeper research on the various network behaviors and the potential patterns isinevitably needed. The spatiotemporal behavior pattern mining on social networkremains an open research direction. It has a large number of research projects still.In this paper, we focus on the research of the spatiotemporal periodic behaviorpattern mining algorithm, based on social network behavior data characteristics. Themain content includes:1. First of all, we propose a hierarchical bipartite graph based spatiotemporalperiodic behavior pattern mining algorithm. This algorithm can mine the periodicbehavior and the moving trajectory of the people in society. It can transform thedynamic social network into a pattern binary tree using the slice of the dynamicsocial network data time, and then obtain the closed periodic location subset whichcan represent the spatiotemporal periodic behavior pattern. The algorithm hasovercome the subset regression problems of the previous algorithm, and it cancomplete the mining task efficiently.2. On this foundation, we put forward a minimum periodic location dominatingsubset algorithm, which can obtain the minimum periodic location dominating subsetapproximately. This subset may cover most individual behavior patterns with leastlocations. We also provide an approximate periodic behavior pattern miningalgorithm. With this algorithm, we can get those approximate periodic behaviorpatterns which are more close to the reality of people’s behavior habit. Then, in thispaper, we design experiments to test the validity of the algorithms we proposed,using the Microsoft Asia Research Institute GPS position data set and the studentmovement trajectory data set. The results of the experiments indicated that thesealgorithms have good accuracy and performance.3. Finally, we design and implement a prototype of the social networkspatiotemporal periodic behavior pattern mining system, base on the MicrosoftVisual Studio2008development environment. This prototype system can analysisand process the input data, and compete the periodic behavior patterns. It can directly output the mining results, related parameters and the statistical data, whichhas contribution to the analysis and research.