Research And Implementation Of Social Network Event Detection Methods

Hot events in the real world will lead to large-scale discussions in social networks,so the analysis starting with social data can reveal the hot issues that people are concerned about.The more people focus on an event,the more frequent users interact with each other in the social network,so it can be concluded that hot events can be identified by detecting dense communities in social networks and digging out keywords that can represent the community in discussion.However,with the continuous expansion of social networks,the traditional community detection algorithm consumes more and more time and space so that it wastes lots of computing resources.This thesis proposes an incremental community detection algorithm based on the traditional K-Clique community detection algorithm,which can effectively reduce the community detection cost in the case of expanding networks.Not only will social networks expand over time,the elements in the network may gradually become ineffective over time.Therefore,this thesis designs and implements the dynamic community detection algorithm based on incremental community detection algorithm for dynamically updated social networks.Both the incremental community detection algorithm and the dynamic community detection algorithm adopt a dynamic update method.When the size of the social network is enlarged,the existing community is updated in an incremental manner.When an element in the network fails over time,the community can be updated in time to eliminate outdated elements in the community.A queue-based TF-IDF(term frequency-inverse document frequency)community keyword extraction method is finally implemented to extract event keywords.By the simulation test of comparing both the incremental community detection algorithm and the dynamic community detection methods proposed in this thesis with the traditional K-clique community detection methods,the test results show that these algorithms can greatly improve the efficiency of community detection with high accuracy.The queue-based TF-IDF keyword extraction method can orderly and effectively extract the keywords of hot events that cause discussion in the social network.