Abstract: Because the existing methods for identifying opinion leaders of the forums are diffcult to capture the dynamic characteristic of network, a algorithm for identifying opinion leaders based on timcvarying graphs was proposed. The algorithm represents the evolution of the network as a sequence of static graphs, each of which represents the aggrega- ted interactions over a given tim}window. Each potential opinion leader is identified by a quantitative indicator, and these results arc then matched each other so that reliable opinion leaders can be identified. Results prove that the pro- posed algorithm has good validity and feasibility.

Key words: Time-varying graphs, Forum, Opinion leaders, Time-window