Abstract: 【Purpose/significance】With the increasing influence of the Internet in society,facing the impact of network mass incidents on social life,it is necessary to grasp the evolution law of network mass incidents,determine the type of event,refine the characteristics of incidents,and put forward targeted countermeasures based on different types of network mass incidents【. Method/process】Through the combination of LDA topic model and K-means algorithm,LDA model is used to realize the recognition of text potential semantics.Finally,support vector machine algorithm is used to perform cluster analysis of network group events,and obtained five types of network group events【Result/conclusion】To build network dynamic identification model of mass incidents, through a large number of text training,in the event the clustering number for 5 good explanatory,completed the network of mass incidents objective classification,are:economy,society,culture form,ethnic group and the environment,provide the basis for the government classification strategy.【Innova⁃tion/limitation】This paper used the LDA model and K - means algorithm,reduce the number of iterations of the model,to determine the best theme,improve the accuracy of the network group incidents recognition results,but the use of text data collected by the wisers news database scope finitely,and the classification results reflect the features of events has certain limitations,further research can fur⁃
ther expand the dynamic text database,improve and deepen the classification algorithm.