With the development of social networks, the interaction between users and the application of social platforms are becoming more and more diversified. The influence and authority of different users are also distinguished in the continuous communication, and this part of the user has a greater guiding force and influence on the public in the process of news dissemination, which is known as the opinion leader in the network. In order to better identify important opinion leaders in the network, it is first necessary to establish a relationship network that maximizes the retention of interactions between social networks. In this model, nodes have influence weights and link edges have link weights, combining these characteristics, this paper improves the LeaderRank algorithm. It is found that the improved algorithm can more accurately identify the nodes with significant influence in the network.