A collaborative model of sentiment analysis and topic mining suitable for public opinion analysis on traffic control from perspective network public opinion was constructed to study the differences in public opinion on traffic control under different types of major public events. In the model, a sentiment dictionary library for public opinion on traffic control under major public events was established and sentiment analysis was performed based on Naive Bayesian classification algorithm. The latent dirichlet allocation (LDA) topic model was employed for topic mining. Taking traffic control under regular public events and emergency public events as the research object, the differences in the changes of network public opinion on traffic control under different types of events was compared by crawling comment data of netizens on the Weibo platform. The public opinion sentiment values of traffic control under regular public events and emergency public events are 0.75~0.95 and 0.35~0.85 respectively. The factors for occurrence of negative topic words of public opinion under different types of major public events are more complex, while the similarity of positive topic words under two types of events is high. Under public emergencies, the popular trend of topic words rise sharply in the early stage and then gradually declined to be flat. The trend of the popularity under regular public events has no obvious regularity.