Research On Overlapping Community Detection And Community Evolution Analysis Method Based On Dynamic Social Network

In the current research on dynamic social networks,overlapping community detection and community evolution analysis are hot topics for scholars.Revealing the community structure in dynamic social networks is helpful to excavate meaningful modules and structural information in the network,and to understand the hierarchical structure and functional characteristics of the network.By analyzing the evolution process of community structure based on community detection results can track the evolution path of community structure and extract important community evolution events,so as to predict and intervene the changing trend of community and even the whole network.For the detection of overlapping communities in dynamic social networks,this thesis first propose an algorithm for overlapping community detection based on label propagation.This algorithm improves on some shortcomings of the SLPA algorithm,introduces the influence ranking of nodes,and fixes the update order of node labels to enhance the algorithm's stability;design a new node label propagation rule,comprehensively consider the effect of label time decay and node similarity for label selection,further enhance the stability of the algorithm and improve the quality of the detection results,then the above algorithm is improved based on the incremental strategy,and the overlapping community detection algorithm of the dynamic social network is proposed.Find the edge set that changes between two consecutive time snapshots,and the set ?V of all nodes associated with it is obtained.Fully consider the range of nodes affected by changes of the network structure,all the nodes in set ?V belonging to the community in the previous snapshot network are added to set ?V,and the nodes in ?V are re-divided to obtain the communities of the current snapshot network,so as to ensure maximize the accuracy of the community while improving the efficiency of the algorithm.For the analysis of community evolution,a community evolution analysis method based on similarity is proposed,and a similarity index that takes into account both community stability and community difference is proposed,and seven community evolution events are defined.Based on the similarity and the definition of community evolution events,the evolution relationship between communities is obtained,and the community evolution events are identified more reasonably.The experiment is divided into two parts:community detection experiment and community evolution experiment.The method proposed in this thesis is run on the real data set and the benchmark data set and compared with the current classic algorithm.The experiment shows that the community detection method in this thesis can get better result of community division,the evolution relationship between communities obtained by the community evolution analysis method in this thesis is more reasonable.