摘要：

Abstract:

【Purpose/significance】In order to enhance the ability to deal with public opinion risk and improve the accuracy of public

opinion recognition and evolution, this study proposes and constructs an evolution model of public opinion risk in emergencies based

on knowledge discovery.【Method/process】Based on the life cycle of public opinion risk and the theory of social combustion, the evo⁃

lution process of public opinion risk is divided into germination stage, diffusion stage, explosion stage and extinction stage from the

perspective of combustion materials, accelerators and fuses. On this basis, the knowledge map is used to represent various definitions

and common sense in the real world, and the semantic knowledge base of association rule structure is constructed, and the convolution⁃

al neural network is introduced as a classifier to obtain the public opinion recognition model. The knowledge discovery method in the

convolutional neural network is used to realize the emotion recognition of public opinions, and the output results are corresponding to

each stage in the evolution process of public opinion risk, so as to show the evolution process of public opinion risk in emergencies

more significantly.【Result/conclusion】Experimental results show that the model has good performance, can well fit the overall evolu⁃

tion of public opinion in emergencies, and has strong practicability.【Innovation/limitation】This study did not consider the heterogene⁃

ity of netizens in the design of the public opinion risk evolution model, resulting in certain limitations in model accuracy, but this also

provides new ideas for further research in the future.