Abstract: This paper is to explore the hot topics in each stage of the public opinion communication cycle of public health emergency microblog, outline the time sequence development trend of the topic evolution of public opinion events, and provide scientific basis for decision-making and analysis of public sentiment.Taking a recent public health emergency as an example, combining the life cycle theory、TF-IDF model and Latent Dirichlet Allocation (LDA) method, the time dimension is incorporated into micro-blog text analysis.The topic mining work includes time series mining of the hidden topic information and the evolution rule of public opinion, and based on the evolution trend of each stage, the corresponding public opinion management and control strategy is proposed.The proposed public opinion evolution analysis method can reveal the discussion content and time sequence development trend of each stage in the public opinion communication cycle.The research provides a theoretical basis and scientific decision support for optimizing the public opinion collection function of microblog platform and assisting relevant management departments to effectively guide and control network public opinion when dealing with similar events.

Key words: Public health emergencies, Internet public opinion, Thematic analysis, Evolutionary analysis, Public opinion analysis, Microblog text mining