Studying on the evolution law of emergency network public opinion, especially the division of public opinion stage, provides the support for government departments to guide emergency network public opinion. Based on the life cycle model of emergencies and the nature law of the evolution of public opinion, the evolution of emergency network public opinion is divided into three stages from a qualitative perspective. Based on the time series forecasting model, this paper builds the temporal model for emergency network public opinion, and analyzes the change of public opinion trend from a quantitative perspective. Based on the Facebook open source framework Prophet, the “Jiuzhaigou Earthquake” dataset was studied deeply and the change point was tested. The rationality of the stage division was verified through the hot word analysis, and the data characteristics of the network public opinion evolution in each stage were summarized. A new idea of combining qualitative and quantitative research is provided for the study of the evolution law of network public opinion.