In this paper, we propose an optimization method based on behavioral data mining for the emergence of negative emotions in network public opinion in unexpected situations leading to problems that endanger social stability and use data mining to confirm the correspondence of the kNN algorithm. In the opinion propagation model, the radius of opinion radiation is assumed to use the network node density and distribution density as features of node exchange information. Then, the kNN algorithm is used to train the comment set for the analysis of user sentiment evolution of network opinion in the social network environment, and the web crawler technology is used to obtain the output interface data APIs of microblogs and WeChat, and the social media user comment data is used as the data for the empirical analysis of network opinion. snowNLP and kNN algorithms are used to analyze the sentiment score of network opinion and the sentiment score less than 0.5, i.e., the sentiment polarity. There were 15 days when the sentiment score tended to be negative and 65 days when the sentiment score was greater than 0.5, i.e., the sentiment polarity tended to be positive.