A Framework Of Community Evolution Analysis In Social Network

In recent years , with the rapid development of the Internet , the Internet has become an important platform of interaction between people,then a social network based on the information has generated . Compared with the traditional social networks, the interaction of people and the model of information transmission are having a lot of change in the social network based on the information. For the study of social relationships in social networks has attracted a lot of attention academia, industry and other researchers,empirical analysis shows that local accumulation is a widespread features in social network, namely community .Community discovery has become one of the important basis research contents in the field of social network.Community-based thesis, the paper does study from the static and dynamic aspects of social networks. First, the paper provides an overview of community detection methods which have been proposed for the static network, and make the analysis of the existing problems. On this basis, combined with the structure hole theory ,a community detection algorithm is proposed ,this algorithm TP-SH make improvements to community discovery method based on topological potential. In this paper, thanks to structural hole considers the number of connections (edge weights) between nodes and the social index of a node which stands for the accessibility of information and the dissemination of information. Therefore, this method can ensure higher accuracy on community detection, and can find overlapping communities.On the basis of the method of TP-SH, a framework of community evolution analysis in a dynamic social network is proposed, which includes social network snapshot, community structure, community evolution path and the analysis of community evolution. In the paper, a tracking community evolution algorithm based on node importance is proposed. In social networks , some nodes have a larger importance value often tend to be stable under the adjacent moment, so in the paper, the text in the next two moments comparing the similarity of the two neighboring communities not only consider the ratio of coincident nodes , but also consider the importance of the node , a new community similarity calculation method is proposed, the method consider two factors, the quantity of overlapping nodes and quality of overlapping nodes, namely the community correlation coefficient . And a new node importance value evaluation method is proposed, the method has high accuracy compared to edge betweenness method,the method has a lower time complexity on the basis of the local shortest path.Over time, an approach to view community evolution is to analyze community changes using a life-cycle model comprising events such as birth, death, split and merge. In the paper,the experimental verification show some important events may be found by an analysis of community evolution path in social network.