Abstract: 【Purpose/significance】To analyze the evolution process of emotional heat of network users in online public opinions caused by emergencies in colleges and universities, excavate the factors influencing negative emotions of users, which is helpful to grasp the evolution law of public opinions in colleges and universities, and assist the administrators of colleges and universities to take effective measures to manage online public opinions in time.【Method/process】A CNN-BiLSTM fusion model is used to analyze the sentiment tendency of microblog comment texts. The value of emotional heat of public opinion is calculated in hourly units. Based on the life cycle theory of public opinion, its heat evolution stage is divided, and then keywords are extracted from the negative user comment texts in different stages. The keywords are analyzed by word cloud analysis and co-occurrence network analysis to obtain the attribu⁃ tion system.【Result/conclusion】Based on two typical cases, a four-level attribution system for user negative comments in campus se⁃ curity emergencies was constructed, and it was found that the first-level attribution elements of user negative comments were mainly internal attribution and external attribution in different stages of user emotional evolution. The secondary attribution elements of inter⁃ nal attribution mainly include empathy and personal situation, while the secondary attribution elements of external attribution include event subject, event analysis and taking measures【. Innovation/limitation】A combination of emotion evolution and attribution analysis methods is used to analyze the reasons for users' negative emotions at different times in a more detailed way. However, due to the lack of elemental weighting analysis, the degree of completeness of the attribution system is lacking.