Social network has been a widely accepted way for people to communicate and interact online. However, few of existing works studied temporal dimension in assessing the authority of nodes on social networks. In this paper, a novel Temporal PageRank (T-PR) algorithm is proposed for analyzing the authority of nodes. Three temporal factors are adopted to personalize PageRank, which favors the nodes that are more important to people. They are Built-up Time-length Factor (BTF), Frequency Factor (FF), and Similarity Factor (SF). The experiments on a real data set demonstrate T-PR algorithm provides the best ranking results over recent competitor methods.