Research On The Method Of Event Diffusion Detection

With the rapid development of Internet technology, the speed of informationdiffusion has become more and more quickly, the method that people acquireinformation has become more and more plentifully. In social networks, a lot ofresearch of information diffusion has been applied to real life, such as websitepromotion, viral marketing, and so on.Due to the widespread of information diffusion phenomena in the social network,the spread of information has become a hot spot of the social network research. But,existing research information diffusion process, little research for the diffusion of theresults of detection.In this paper, after the analysis of the traditional diffusion model LT and ICmodel, encapsulate the information in the diffusion process, given complete definitionof diffusion events. And describes the concept of the event, event graph, diffusiongraph that are used in the diffusion events. A large-scale graph of the vast amounts ofinformation was formed after the information diffusion, will be based on thelarge-scale structure, detecting diffusion of similar events. Given the decision of thesame type of event in standard and large-scale structure of a graph partitioningstrategy. Information diffusion process, because it contains a mass of information, andthe diffusion is not timed to occur, and thus pre-process data before the operation ofdetection for event information.Next, a distributed event detection algorithm was proposed based on theMapReduce programming model. Diffusion events detected after diffusion detection,and then proposed a distributed feature detection algorithm. Analysis of Graphstructure feature of diffusion events, to divide feature for diffusion events.Finally, by experimental results, observe the implementation efficiency of thedetection algorithm, quality of detection. And analysis of results to event detectionalgorithm on different detection parameters. Reasonable range of values givendetection parameters, if detection parameters take values within the range, then diffusion events can be effectively detected, show that the algorithm has broadapplicability in real life.