Research On Community Detection And Influence On Information Dissemination Based On Structural Analysis

With the rapid development of the Internet and the increasing popularity of intelligent terminals,online social networks have become a vital approach for people to access information,disseminate message,make friends and entertainment.Huge amounts of users in online social networks established their own social relationship network by taking the way of following each other.By utilizing the manners of posting,forwarding,reviews and recommendations,information in an online social network along the relation links between users to carry out fission dissemination.Commonly,online social networks not only present a general structure of community,but also provide a channel for information dissemination and affect the spread of information.Emerge as the issues require,we take the structural analysis of social networks as the research focus in this paper,which includes mining the communities in social networks,constructing information dissemination tree(IDT)by tracking the information dissemination paths,and the topological characteristics of information dissemination tree are also analyzed by using null models and then further indirectly research information dissemination properties and the structural influence on information dissemination.The contributions of this paper are summarized as follows:(1)An unified framework for simultaneously detecting overlapping and hierarchical communities from social networks is summarized.Under the guidance of that framework,a maximal clique based overlapping and hierarchical communities detection algorithm is proposed,in which the “coupling strength” function for the community similarity calculation as well as the communities quality assessment function based on the community density are also proposed.The experimental consequences on serval benchmark social networks,which including Zachary,Dolphins and Coauthorship,indicate that the proposed algorithm performs well on the consistency between detected communities and real community partition of a social network.Moreover,compared with the algorithm of EAGLE and CPM,and the better performance in the term of accuracy of community detection is obtained.(2)Due to online social networks have the feature of strong sparsity,the number of maximum clique is very small or even nonexistent.However,in the case of without considering the individual nodes and edges in a network,each node is either located in a closed triad,or located in an open triad.Based on this observation and also serve as a supplement to the maximum clique based community detection algorithm,a triad percolation method for mining community from a social network is proposed.The experiments on several benchmark social networks indicate that the found communities keep consistent with the corresponding community partition of a network.Meanwhile,compared with the CPM algorithm,the results illustrate that the TPM algorithm proposed in this paper can better find the community structure in the social networks with strong sparsity and has a higher accuracy of community detection than the CPM algorithm.(3)The online social networks provide many channels for information dissemination.For the purpose of exploring the information dissemination rules and the influence of network structure on information dissemination in SinaWeibo,IDTs are constructed by tracking the information diffusion paths,which are dendrograms composed of open triads.Then,taking into account the occasional nature of information dissemination in social networks and the limitations of the data sample,we use the null model based on null hypothesis in statistics and extend the traditional significance profile to analyze the structural characteristics of SinaWeibo IDTs,which include the degree correlation,information dissemination path cascade ratio and the influence of community structure on information dissemination,and thus indirectly study the information dissemination is SianWeibo.The empirical analysis and experimental results on Sina IDTs show that community structure also commonly exists in the Sina IDTs and has the linear modularity distribution,while information is widely spread in the community and its dissemination among communities is suppressed;The connections between the nodes with small out-degree of Sina IDTs and the large out-degree nodes are strongly suppressed,which have the function of hindering the information dissemination,while the connections between the nodes with small out-degree and the nodes with larger out-degree,as well as the connection between the nodes with small out-degree and the nodes with moderate out-degree are weakly suppressed,which have the promotion function to the information dissemination.Moreover,the comparison results between the classical Price model and the IDT null proposed in this paper verified the feasibility of the proposed IDT null model in the analysis of information dissemination,and the IDT null model also solved the problem that the Price model can not directly generate the random copy of the information dissemination tree.(4)On the basis of structural analysis for Sina IDTs,we introduce the time-stamp characteristic into the nodes and edges in Sina IDTs,and then study the relation between network structure and information dissemination burstiness.The corresponding experimental results on Sina IDTs indicate that the structural properties of SinaW eibo network make contributions to the explosive dissemination of information,and the '5-hour effect' of information dissemination in SinaWeibo.