Research On Community Detection Algorithm Based On The Measure Of Set Pair Similarity

Community structure as an important property of the social networks can reflect the social individuals’ behavioral characteristics and the characteristics of the relationship with other individuals. Community detection is helpful to the analyses and understanding of social networks’ internal regularity. And it has a very important theoretical significance and research value on the social network research.This paper considers the set pair analysis theory in the discovery of social networks’ community detection.And then we propose a measure of the similarity between vertices which is on the basis of the set pair connection degree.And we use the measure to detect communities in social networks.Firstly, according to the description of sets’ same, uncertain and opposite relations in the set pair connection degree, we define the measure of similarity between vertices which based on the relation connection degree take into account weight and clustering coefficient to improves the accuracy of measures on vertices’ similarity. The measure considers the influence of variant mark i on the transformation of uncertain relation into same relation.We take the clustering coefficient of vertex as i. Considering the contribution of vertex degree and paths between vertices to the similarity, the measue give weight to same, uncertain and opposite relations between vertices.Secondly, apply the measure of the similarity between vertices to community detection. A hierarchical clustering algorithm VSFCM based on the combination of vertices similarity first cluster and communities mean cluster is proposed.to avoid a large number of update operation in the traditional hierarchical clustering algorithm and to reduce the unreasonable community clustering.Thirdly, the existing methods of social network analysis based on set pair analysis theory only treat traditional unsigned networks as research objects.Considering the features of signed networks, including the network topology and symbol attributes of edges, we define the measure of similarity between vertices in signed networks which based on the relation connection degree of set pair. Combining the structural balance theory and clustering coefficient of vertex, we assign a value to i.Finally, experiment on traditional social networks and signed networks to detect communities.The results show that the definition of similarity between vertices is rational and the algorithm VSFCM is correct and effective.