In a blogspace, citation behaviors reflect interests of bloggers. To fully get insight into the latent information in a blogspace, in this paper, we ntend to mine popular co-cited communities consisting of core sets and follower sets. In such a co-cited community, bloggers in the core set are frequently cited by bloggers in the follower set and the co-citation behaviors among bloggers are very intensive. Through co-citations, not only the popular core-set nodes but also the followers can be discovered. As such, one could effectively obtain the trends of discussion among bloggers. Explicitly, two kinds of co-cited communities are exploited: perfect co-cited community and approximate cocited community. Given a blogspace, we first transform this blogspace into a transaction database. Then, by exploring frequent closed itemset mining, we are able to discover perfect co-cited communities. Then, a greedy algorithm is proposed to derive approximate co-cited communities. To evaluate our community structures mined, we conduct extensive experiments on deli.icio.us dataset. The experimental results demonstrate the effectiveness of our proposed framework.
