Researches On The Node Role And Community Detection Of Weighted Social Networks

Social network is one of the basic paradigms used to express the relation between social members.Because of the structure of social network is becoming more and more complex.we usually model social network with complex system.Community structure is another major characteristic of complex network after the discovery of small world and scale-free characteristics,the research on community structure is of great help to understand the network structure.However,the existing community detection algorithms mainly focus on the network topology,while ignoring the weight of lines with actual significance.Firstly,this thesis use the CPM algorithm on several real datasets for community detection,finding that it is of highly dense for the internal connections of communities detected by CPM algorithm.Communities with highly dense internal connections can be regard as the core architecture of communities.Basing on the core architecture,we obtain the core community,and then put forward the community detection algorithm based on the core-periphery theory for the unweighted networks and the weighted networks respectively,by combining the Core-periphery theory and the Two-step flow hypothesis.After that,we compare the new algorithm with other algorithms on real datasets including weighted networks and unweighted networks.The experimental results show that the algorithm can effectively find the community structure.Due to the complexity of the network,nodes in the network mostly play different roles.It is the core nodes in the core of the network and bridge nodes that transmit information catching more attentions.Detecting these important nodes and analysing these nodes is of great actual importance.We argue that core nodes must show in core communities because of their core position,and then propose the core nodes recognition algorithm based on core communities.Furthermore,because of the information role of bridge nodes,bridge nodes must exist in the edge of community and transmit information between nodes of core communities.Basing on the theory,we propose the bridge node recognition algorithm.Then we do experiment on five real networks including weighted networks and unweighted networks,and compare with the result with other algorithm.Results show that the role recognition algorithm proposed by us is currently effective.