A conventional model of public opinion analysis is no longer suitable when the internet is the primary arena of information dissemination. Thus, a more practical approach is urgently needed to deal with this dynamic and complicated phenomenon of propagating public opinion. This paper proposes that the outbreak of internet public opinion and its negative impacts, such as the occurrence of major security incidents, are a result of coupling and the complex interaction of many factors. The Functional Resonance Analysis Method model is composed of those factors and considers the stages of network information dissemination, the unique propagation rule, and textual sentiment resonance on the internet. Moreover, it is the first public opinion governance method that simultaneously highlights the complex system, functional identification, and functional resonance. It suggests a more effective method to shorten the dissipation time of negative public opinion and is a considerable improvement over previous models for risk-prediction. Based on resonance theory and deep learning, this study establishes public opinion resonance functions, which made it possible to analyze public opinion triggers and build a simulation model to explore the patterns of public opinion development through long-term data capture. The simulation results of the Functional Resonance Analysis Method suggest that the resonance in the model is consistent with the evolution of public opinion in real situations and that the components of the resonance of public opinion can be separated into eleven subjective factors and three objective factors. In addition, managing the subjective factors can significantly accelerate the dissipation of negative opinions.