Research On Community Detection Algorithms In Complex Social Networks

With the new century of complexity and networking,many complex social issues can be described in the form of networks.Therefore,the analysis of complex social networks has become an important issue for related scholars.In complex social networks,if people are abstracted into nodes and the various relationships are abstracted into edges,the complex networks will be formed with nodes and edges.Scholars have found that many actual networks have similar structural characteristics: the community structure.The community structure shows the characteristics that the internal nodes of the community are more closely connected and the connections between the communities are sparse.The community detection is an important method of excavating the network community structure.It is important for network division,scientific research,and marketing to find the community structure in social networks.In order to solve the shortcomings of low accuracy and effectiveness of some classical community detection algorithms,this thesis proposes a community detection algorithm based on expansion of the central node.In the method of selecting the initial central node,the algorithm firstly considers the importance of the node degree and normalizes it as the degree weight of the node.It also considers the edge aggregation coefficient that can express the characteristics of the community structure,the sum of the degree weight of the aggregation coefficient is multiplied to obtain the final initial central node,and then the community expansion is performed according to the local fitness function.Finally,the algorithm is simulated on the Zachary and Dolphin public social network datasets.The experimental results show that the algorithm has good accuracy and validity,and it can also discover overlapping communities.The community detection algorithm based on expansion of the central node commendably solves the problem of community detection in structural ne tworks,but it may not be applicable to hot social networks because these hot social networks are not only a single structural network,they also hide basic attribu te information between members.Therefore,on the basis of the above algorithm,this thesis analyses the characteristics of basic attribute information of members in social networks,and then proposes a community detection algorithm based on parallel expansion of central node.The algorithm firstly proposes a static influence calculation model based on Analytic Hierarchy Process and a dynamic influence calculation model based on Page Rank algorithm,then it proposes the method of calculating the centrality of the node,and the central nodes in the network can be excavated according to the centrality of the nodes.The parallel community expansion is performed according to the local fitness function.Finally,it is proved feasibility and accuracy by the experiment s with crawled Sina microblog data.