Analysis of Public Opinion Evolution in Public Health Emergencies Based on Multi-fusion Model

This paper analyzes the public opinion topics and the emotional fluctuations of netizens during the closure of the city due to the new crown epidemic, and reveals the correlation between public emotional fluctuations and public opinion topics during various periods of public health emergencies. We use the BERT-BiLSTM fusion model to efficiently capture the two-way relationship in the sentence and improve the accuracy of sentiment classification of Weibo text at the same time, to extract the hot topic feature words at different periods of the city closure by LDA mode; finally, the SEI7R model is used to simulate the prevention and control recommendations of various periods of public opinion proposed in this paper to verify the effectiveness of the prevention and control recommendations. The experimental results show that: The F value of the BERT-BiLSTM fusion model in the classification of public sentiment polarity can reach up to 0.907, which can effectively classify the sentiment of netizens; The LDA model can effectively dig out over time The theme characteristics of the text that have gradually evolved over time. The simulation results of the SEI7R model show that the countermeasures and suggestions put forward in this paper can effectively provide a theoretical basis and methodological reference for public opinion management of public health emergencies.