Abstract: The relationship between user profiles and the data of microblog content in Sina microblog was obtained by programming and web crawler, and a variety of explicit or implicit relationships between microblog users were discovered by using data mining. On the basis of this, a semi-supervised user interest matching prediction algorithm was proposed. According to the individual state division method of compartment model, a network public opinion propagation model is constructed based on user interest matching through state transition analysis and inference of state transition probability. The results show that the model can well describe the laws of public opinion propagation in social networks, and reproduce the real propagation process of network public opinion in the social network from the perspective of complex networks.