Research On Community Detection And Centrality Analysis Algorithm Based On Social Networks

The social networks develop fast, widely used in various fields, such as science and technology, commercial, economic and biological area, which shorten the distance between people, changed the ways of people’s learning, working and living,affected the social patterns. The huge amounts of information on social networks can represent the behavior of the social person or organization. These network data can be used to analyze the behavior of the individual or group, can be effectively used in the research of community detection, centrality analysis, data classification, interest recommend and other aspects. And the community detection and centrality analysis based on social networks are two important research topics, which have become hot research in recent years.Community take the form of clusters. There is a very strong interaction between the internal nodes,and community detection is to identify the different clusters in social networks. Community is widely exists in the field of biology, medicine and the Internet, so it is meaningful to do the research of community detection. The importance of individuals or organizations in social networks is different, some nodes will be in a more important position, play an important role in the collective. We use centrality to describe the importance of nodes. It is helpful to do research on centrality analysis based on social networks, and it does good to public opinion analysis,marketing, web search and other aspects, also it has a high value on both research and practical application.Based on the analysis of previous algorithm, we propose a community detection algorithm based on rank centrality called RCCD(Rank Centrality Community Detection). We propose the concept of rank centrality in RCCD algorithm, and use itdo the ranking of the nodes in networks. We select the front nodes as the initial node sets. And based on the idea of PAM clustering algorithm, we use the represent object as the center node to divide the community. Firstly, we use the method of eccentricity centrality to rank the nodes, and select the front nodes which has threshold constraint in the initial seed nodes selection, the distance of which is as far as possible and dispersed as the initial k node sets. Then use the cosine method to calculate similarity between nodes, and divide the nodes in the networks into different communities according to the similarity principle. Then based on the idea of PAM clustering algorithm, doing the replacement of represent object(center node).In the replacement,we use the method of eccentricity centrality to rank the nodes, thus forming the center nodes candidate set. Then select a node from the center nodes candidate set in turns to do the replacement, thus updating the center node. Algorithm iteratively until the center nodes do not change or the structure of community is no longer changed. We take experiments on Karate Club datasets and American College Football datasets,and compare the results with other algorithms, verifying the effectiveness of the algorithm of RCCD.In addition, we proposes a kind of centrality analysis algorithm based on the subjective bayes algorithm(CAB). We choose three kinds of traditional index to measure centrality. They are degree centrality, betweenness centrality and eccentricity centrality. We use them to build the uncertain evidence synthesis index set, finally combining the subjective bayes method to do the synthesis, thus getting the ranking of the nodes’ centrality. We take experiments on Karate Club datasets and American College Football datasets, and compare the results with the method of using only one index to do centrality analysis, verifing the effectiveness of the algorithm of CAB.