A Research Of Community Mining In Social Network Based On Granular Computing

Social network is one of topical complex networks, used to describe therelationship between the members of society. With the development of networktechnology and mobile devices, communications between people become diversity,and the social network sites spring up like mushrooms. Community structure, whichis the focus of recent studies on social networks, is an important structural feature ofsocial networks. It is a series of entities connected closely to each other, the nodes inthe same community structure have closer relationship than the nodes in differentcommunity structures. The relationship between people as well as the communitystructure formed by it are the core of social network.Therefore, it is of greattheoretical and practical value to study on complex network theory specially on thecommunity structure of social network.The research begins with the community structure in social network. Firstly, thepaper introduced the concept of the complex network and granular computing, andfew existing data mining algorithms. Then with the idea of granular computing, thepaper proposes a node similarity measure based on network topology to measure thesimilarity between nodes and communities, also between communities. According tothis, two efficient algorithms for detecting community structure are given, namelyCommunity structure Mining Algorithm based on-quasi Complete sub graph andCommunity structure Mining Algorithm based on Granular Computing. An actualexample associated with the real social network dataset of Zachary Karate Clubclearly presents all steps of the algorithm. Finally, in comparation with GN algorithmand CNM algorithm, the two algorithms are tested on three real social networks invarious scales. The results indicated that the algorithms proposed in this paper aremore effective. They can get community structures with higher quality. The twoalgorithms are depicted as follows:Community structure Mining algorithm based on-quasi complete subgraph(α-CGC Algorithm):The core of this algorithm is to use a-Quasi complete subgraph given by this paper as the center of the cluster initialization.Then,in order to get the finalcommunity structures, a coagulation progress is handled through the closerelationship between node and community, also between communities. It is indicatedthat the community structure detected by this algorithm will not be influenced bydifferent initial nodes.Community structure Mining algorithm based on Granular Computing (CGCCAlgorithm):The core of the algorithm lies in designing a granularity rule on networkstructure using granular computing based on rough set, as well as generating anetwork granularity space under it. Then, a model of detecting community structurewith granular computing is proposed, which converts the problem of detectingcommunity structure to problems solving on different granularity spaces.