Abstract: Taking the mass topological data at the Internet AS (autonomous system) level provided by CAIDA as sample data, the sample coverage and prejudice in sampling are discussed to correct the sample data. Then, the topological data at the Internet AS level from 2001 to 2007 are divided into different communities, and the eigenvalues relevant to community evolution are calculated, such as the modularity and community size. The results showed that the topological data at the Internet AS level exhibit more and more apparent characteristics of community gradually. Simultaneously, the dynamical behavior of the nodes in the AS domain is analyzed according to the eigenvalues relevant to community evolution, and what causes the evolution is analyzed as well.