Research And Application On Topology Analysis Of Online Social Networks And Community Detection

With the rapid development of economic and social, online social network plays a more and more important position in people's ordinary life. Online social network become power and power with the continued development of Internet and mobile Internet. It is one of the hotspots in the research by analyzing and mining the social network data to give hands to people 's decision-making action. This thesis pays attention to analyzing the topological characteristics of online social network and the community discovery. The main work of this thesis can be described as follows.Firstly, the preliminary concept and topological characteristics of online social network are summarized. Then, this thesis discusses the properties of each effective node. By degree,degree distribution, average path length, clustering coefficient and other statistical methods,the nature and topology, and so on, of social network can be obtained. We also study the structure of social network, especially on topology.What's more, based on the comparison and research of traditional algorithms FN and NOCDI, an improved community discovery algorithm named ICDI based on modularity increment is proposed. By initializing the structure of online social network according to the modularity increment concept and calculating the increment of the modularity, the largest increment of modularity in online social network can be obtained at the same of searching of the network with the theory of graph theory. Then, multiple communities selected should be merged on modular increment decreasing. Next, the entire online social network will be updated according to the structural characteristics of it and loop once more iteratively.Experimental results demonstrate the advantages of ICDI compared with FN and NOCDI.Finally, in view of the inherent shortcomings of the existing social network community discovery algorithm focused on small-scale online social networks, this thesis proposed the algorithm for community discovery by method of clustering integration on the community structure of online social network. By community dividing with AP, each cluster can be considered as a single community. With different reference values once again appending with AP, different results can be obtained. Then, the number of communities can be obtained by voting algorithm. The results of AP are regarded as the base of clustering followed by clustering of CSPA, HGPA and MCLA. In this way, community discovery with better effect is obtained. The experimental results validated the accuracy, feasibility and rationality of community discovery algorithm proposed from clustering coefficient and others alike. The community discovery algorithm proposed by us can achieve better result in community discovery on online social network. The research of this thesis on online social network community discovery shows a new strategy for other study on online social network and will obtained more and more attention.