By abstracting the complex structure of social network into undirected graph, nodes into pages and edges into hyperlinks, and combining PageRank algorithm with the discovery of key nodes in social networks, this thesis comes up with a new algorithm of key nodes premised on improved PageRank algorithm, and finally employs microblog data as data set. Through the experiment of comparing KeyRank algorithm with TIPR algorithm, it can be concluded that the PageRank algorithm proposed in this thesis is comparatively suitable for discovering key nodes. Under the same data set, the efficiency of identifying key nodes is raised by 30% in comparison with the other two algorithms.