This paper mainly uses the microblog "Didi Wenzhou Girl Killed" topic as the data source,calculates the emotional value of comments,labels the positive and negative emotions,constructs the emotion classification model through Word2 Vec and SVM,and uses the Word2 Vec method to calculate the words with high similarity to the five objects(didi,drivers,customer service,girls,police) involved in this event,and discuss the emotional trend from two aspects:emotional time series analysis and emotional evolution analysis about public opinion objects.It is found that the sentiment classification model can effectively predict the emotional trends of netizens;the change of the time series of netizens’ emotions coincides with the evolution of public opinion;the Word2 Vec word similarity calculation model can effectively reflect the emotional attitude of netizens to five objects and topical features in the stage of public opinion.