Research On Technology Of Community Detection In Online Social Network

The lifestyles of people is being changed with the development of computer network technology during the last two decades.Due to the widely use of Web 2.0,the Online Social Networks are burgeoning.The social network has an important structural characteristics—“community”,which implies that the network is composed of a lot of groups and nodes from the same group connect closely while nodes from different groups connect loosely.Also,nodes within a same community may share a common nature,and hence play a similar role in the network.Discovering the network structure not only helps to understand the function of network,but also is helpful to identify the connect level within a complex network.As a result,this research is of importance from both theoretical and practical perspectives.Our major work are as follow.For the reason social network node has its own attribute,this paper proposed a community attribute entropy combined method based on existing community detection method,which makes detected communities meaningful.For the reason,original label propagation algorithm is not robust,this paper proposed a new method with new update strategy,which improve the robustness of label propagation algorithm with limited time upgrade.Label propagation algorithm is an effective method on large-scale complex network community discovery,but the robustness of the problem has not been solved.This paper also propose an improved algorithm named NLPA based on label propagation algorithm to enhance its robustness.The algorithm takes into account the importance of the nodes to be updated and it is included in neighbor maximum frequency tags statistics,the algorithm also prevent premature formation of ultra-large-scale community by controlling the threshold.The algorithm is used to analyze Sina Weibo,which demonstrates there is a strong relationship between structure features and semantic features in social networks.