Research On Community Discovery Method Based On Theme And Structure

With the development of technology,people realize that real-world networks are not regular networks,and they are different from random networks.They have many complex features like small-world,scale-free,self-similar,self-organizing and so on.A significant shared attribute in a complex network is the community structure,that way to say,internal nodes in the same community are tight linking but the nodes between the communities are sparse.Community discoveries contribute to us to detect the internal topology of the networks and discover the community structures of networks and hidden rules.It has very important theoretical and practical value to further understanding of network functions and predicts network behavior.Community structure discoveries research has been widely used in personalized recommendation,advertising,information retrieval and other fields.Also it is one of the hotspot of researchers who are in the field of complex networks.This paper focuses on the community discoveries in complex networks and discuss in the following three aspects.(1)A community discovery algorithm SSMCA based on structural similarity is proposed.Based on the idea of structural similarity,this algorithm is used to solve the problem of low accuracy of existing community partitioning algorithms.Structural similarity is regarded as the edge weight.The edges in the network are sorted in ascending order according to the weight,and the edges with lower weights are deleted to form several isolated communities.Then merge the main communities according to the structural similarity and module difference between communities.Quantitative evaluation method is used as iteration condition of the algorithm to obtain high-precision community structure.Experiments on artificial benchmark networks and real data sets show that the performance of this algorithm is better than other classical methods in the accuracy of community partition.(2)A topic-oriented method to community discovery is proposed.Communities that the method discovers are not only reflecting the strong and weak relationship of the connection structure,but also unearthing the common theme within the communities.First,this method proposes a high-impact user evaluation algorithm HIUEA,which is used to discover high-impact users in the networks.Then dig out the behavioral content of the common users related to it and perform text clustering through the EWKM algorithm to obtain clusters of social objects clusters containing the meaning of the topic.Next,using the SSMCA algorithm for different topic clusters to topological analysis,and finally getting a close-knit and single-topic community structure,following on validating on Zhang Jie's fan networks and the university patent cooperation network dataset.The results show that the community structure which has been discovered by this method can effectively reflect the topology and semantic information,making the community structure more meaningful.(3)Based on this paper which proposes the topic-oriented community discovery method,the Weibo theme recommendation system is designed and implemented.The system integrates high-impact user evaluation algorithm HIUEA,classic text clustering algorithm EWKM and similarity-based community discovery algorithm SSMCA.It achieves to dig high-impact users and analyzes the connection structure to obtain related topic fans data.