The event knowledge graph of emergency was constructed to reveal the logical correlation and evolution path between public opinion events, and provide reference for emergency analysis and response. Firstly, based on the rule template, the logical relationship of events in the network public opinion of emergencies was extracted, and the event knowledge graph was constructed. Secondly, the events were vectorized and generalized, and the K-Means++ algorithm was used to cluster and abstract similar events to build an abstract event knowledge graph of network public opinion. Finally, the target events were deduced and predicted on this basis. The results show that the types of event relations in the emergency event knowledge graph are mainly causality, coordination and sequence, and most of the events have short logical chains and no obvious correlation with other events; The construction of the emergency event knowledge graph is conducive to analyzing the evolution of the emergency, grasping the key event nodes and predicting the development direction of the events.