The Study Of Community Discovery Method On Large-scale Social Network

Social network is a combination of numbers of individuals,or organizations,as well as their connections.In recent years,as the rapid development of the internet and social network sites,the study of large-scale social network has become hot.In large-scale social networks,some sociology features,such as small world and so on,will become more obvious.So the study of large-scale social networks can help us better understand the social relations of everyone.This thesis proposes an approach to community discovery on large-scale social networks,based on the features of social networks,such like centrality,small community, small world etc.Firstly,the communities at the bottom level are elected from the original network used the mutual connection;next,these bottom communities are regarded as one node to reformat the network,at the same time the weights of edges are calculated and added to the new network;at last,clustering method based on local-modularity are employed to find the communities.How to evaluate and interpret the communities is another important issue.This thesis tries to use statistics to explain the founded communities,labels every community found by our solution on KDD-Cup dataset.Such approach can be applied to the analysis of large-scale online social network communities.The experimental results show that the scale of the communities that our method found,is more well-proportioned,and most of these communities are relative small.In fact,the small community is stable and easy for interpretation,therefore the discovered communities will be more rational and reliable for the future potential use.