With the rapid development of virtual communities, lots of knowledge has been expressed or produced many times. This paper focuses on how to collect and organize the knowledge in communities, with the theory of social network. According to the existing data of affiliation, there are numerous relationships between the questions and their answers in the community. We take “Zhihu” for example, and build the net model, which could avoid the problem that the connection between notes is based on subjectivity. Meanwhile, based on the cohesion of social network, we put forward the concept about “knowledge subgroups”. The result shows that as an effective analytical method, social network could explore the inner connection and similarity among the notes in knowledge network, thus achieve the aim of knowledge discovery and classification.