Community Discovery Method Based On Users’ Interest And Network Structure

Along with the boom of the network, Online Social Network developed fast in recent years. As a both new social media and social network tool, Microblog has captured the favor of the public for its characteristics of’short, flat and fast’Firstly, the paper summarizes the results and looks forward to the next research. There is0.3billion people using it in our country, which is half of all internet users nearly, ranking first in the global. In the case, not only administrators but also users are facing with many opportunities and challenges. How the users find friends in the boundless; Can the companies find interesting information from the big data; How the administrator manage the huge network et al. Community Mining can help to solve these problems. Community is a group of similar users. There are many researches which based on the complex network. However social network consists of human beings and links between them, which is different from the traditional one. In this case, the content of nodes cannot be ignored. So, the paper aims to find a method considering content of the nodes and network structure.First, the paper does a research on the social network, analyzing the feature of behaviors of users through forming models of users-the influence model and the theme model. The former can identify the core users who have a strong appeal to the others in the network. At the same time, we think that feature of users can be found from its words, so the latter will extract their themes from their publishing microblog. On the basis, putting the core users as the initial cluster nodes, this paper establishes the clustering model, which can not only improve efficiency but also avoid the local optimal results. Then cluster the users based on their similarity, taking their distance on the network in account, to ensure that one community consists of the users who are both similar and close. And merge the small communities to make the results owe more application values. Finally, we applies the method on the real data, It finds that the method can not only find out the potential communities, but also get the theme of communities, solving the problem that the traditional way couldn’t explain the results semantically. In addition, the paper makes a summary of the results, and looks forward to the research work in the future finally.