A Rumor & Anti-rumor Propagation Model Based on Data Enhancement and Evolutionary Game

Abstract

In social topics, the study of rumor propagation mechanisms is essential to master social behavior. In this paper, aiming to explore the adversarial game relationship between rumor and anti-rumor in the propagation process and considering the sparse of real and effective data in the process of rumor propagation, we propose a rumor propagation model. First, considering the sparse data samples of effective cross-countermeasures in the process of spreading rumor and anti-rumor, a model of unsupervised GAN is used to enhance homomorphism data in the sample space. Second, aiming at the symbiotic and antagonistic game relationship between rumor and anti-rumor, evolutionary game theory is introduced. The internal and external factors that affect user behavior within rumor propagation are comprehensively considered, and a quantitative rumor-refutation mutual influence model is proposed. Finally, considering the dynamic time limit of rumor and anti-rumor propagation life cycle, we herein propose a simple prediction method of rumor topic group behavior based on a dynamic iteration mechanism. Experiments show that our method gets 9%-29% precision gains over the current methods on different datasets.