Research Of Hierarchical Community Structure In Complex Network Based On Maximal-Degree Node

Social Networks Analysis is a subject which uses quantizing method to analyze social networks characteristic and the individuals properties. With analyzed from the regular network, random network, to the current complex network, more and more matter from different areas can be shown as complex network,so that one is impossible to learn global data information in real-time due to mass data and dynamic network structure. Therefore, local community detection has been a hotpost this years.Local community hierarchy detection studies the social hierarchy that each local communitiy existed, which is different from the community detection methods in the past . By analyzing the spatial structure of community, we can better know the node's status in network,understand the community structure and its evolution, and also the community that belongs to which hierarchy of network. What's more, there is a wide range of applications from community hierarchy detection, such as user behavior analysis, anti-terrorist tracking, etc. Most hierarchy community detetion methods found recently are based on the existed community detection algorithms, but some of them have limitations: lack of effective evaluation methods for detection suitable community structures, and also fail to find the sub-communities structure in certain hierarchy.To solve the problems above, a new algorthim based on the local community detection method and maximal-degree nodes for aggregation properties, is proposed to discovery the community hierarchy: Hierarchical Community Strucure Detection based on Maximal-degree Node (HCSD-MN). First,the algorithm has a new perspective to social network: as one start node associated with one or more maximal-degree node(s), there are also different topic communities associated with these maximal-degree node(s). Second, the start nodes set is generated by start node and its neighbor nodes that have high similarity. Third,With adjusted the community-resolution parameter, the evolution of community hierarchy structure would be shown.After that, the hierarchy bound would be found according to the relationship between community resolution and community links density ratio,while the research also obtain the community structure saturation by hierarchy bound .In this paper,NMI is used to evaluate the result of HCSD-MN algorithm,which is also compared with other algorithms.NMI is a criteria widely used for community hierarchy detection methods. Experiments show that our algorithm can not only improve the result of community structure detection, but also successfully detect the hierarchy bound and the sub-communities structure in certain hierarchy of community .