Modeling public opinion dynamics in social networks using a GAN-SEIR framework

It is often accompanied by the spread of network public opinion events when an emergency occurs, which is easy to cause obvious emotional fluctuations in society. Therefore, how to build a more realistic evolution model of public opinion, so as to grasp and predict the development trend of public opinion in time, is an urgent issue. Based on the Generative Adversarial Network (GAN) and SEIR model, this paper constructs the GAN-SEIR model for the simulation of the evolution of public opinion in social networks. Firstly, an evolution model of network public opinion is constructed by referring to the SEIR epidemic model. Secondly, based on the generative adversarial network, the relationship between the interaction of each element in the system is determined, so that the information propagation in the GAN-SEIR model is more consistent with the complexity of actual propagation. Then, the system dynamics module of Anglogic platform is used to simulate the evolution trend of public opinion. By simulating the logical structure of each related element in the network public opinion system, the dynamic feedback structure and mechanism of the influence factors of each element in the system are analyzed, and the future development trend of the network public opinion is obtained from the internal motivation of the public opinion event system. Finally, the actual data of China Eastern Airlines 3.21 accident was used as a real case support for verification. The results show that the error between the predicted data and the actual data is within 3%, which proves that the model can effectively predict and track the development of network public opinion.