Abstract: With the development of social network,it becomes more and more important to extract useful information from the social network and provide valuable knowledge to users in an interactive visual interface intuitively.Clustering,as a crucial method in data mining,offers the global data analysis results.Traditional clustering methods of social network data mainly consider network topological structure.However, they haven’t considered the user interests for clustering.In this paper,the users are clustered by computing user-interest similarity based on Bayesian probabilistic model,furthermore,the interactive visualization method is designed to present the user clustering results.Specifically,we computed the feature vectors representing users’ interests based on latent semantic model.Then clusters with different interest characteristics were built based on these feature vectors.The suitable number of clusters are determined by heat map visualization results.Finally,we presented the interactive visualization method based on hierarchical bubble chart to support users to explore the clustering results from the global overview to local details.We performed experiments and analysis with data crawled from Douban website.The results validate the effectiveness of our method.

Key words: Social network,Clustering,Data visualization,Latent semantic model