Abstract: 【Purpose/significance】Network public opinion security is an important component of social security. Weibo, as a special social group, not only contains a large amount of short text information, but also various non textual information interwoven together, re⁃ flecting the formation and evolution process of hot topics in network public opinion. How to identify hot topics from multi-source het⁃ erogeneous texts and predict the development trend of public opinion events has become a focus of current public opinion research. 【Method/process】Using hypernetwork theory to model sudden network public opinion events, constructing hypernetwork models of subject subnets, object subnets, information subnets, temporal subnets, and emotional subnets, applying social network analysis, deep learning, and Pagerank algorithm to calculate the viewpoint weight of network public opinion, and identify key nodes of network public opinion【. Result/conclusion】The research results indicate that different subnets have a significant impact on the evolution and devel⁃ opment of network public opinion, and social network analysis algorithms can accurately measure the key nodes of network public opinion.【Innovation/limitation】Applying the theory of hypernetworks to construct different subnets for analyzing hot topics in network public opinion, predicting the evolution trend of network public opinion, and expanding the research sample in the future is necessary to make the research results more universal.