In social network analysis, generating random graphs is still an important issue. In this thesis, we propose one new generative model for one-mode social network in which separating components, popular individuals and transitivity e ffect are considered. We also implement the bipartite social networks and BA model in order to compare their features to our model. The simulation results show that there are really signifi cant di fferences. The new models also explain some interesting phenomena in real-world social networks including the power-law degree distribution, the small world network and so on.