A social rumor and anti-rumor game diffusion model based on sparse representation and tensor completion

In view of the high dimensionality and complexity of the rumor diffusion space in social networks, and considering the game confrontation in the process of rumor diffusion, this paper proposes a rumor and anti rumor game diffusion model based on sparse representation and tensor completion. First, taking advantage of the sparse representation's ability to represent all of the original samples with as few atoms as possible to enable low-rank vectorization of the feature space. Second, in view of the fact that a tensor complement can recover lost data with high precision, a tensor complement with a time-decay function is used to complement the dynamic behavior data of users during the rumor process. Finally, considering the game relationship between rumors and anti-rumors during diffusion, evolutionary game theory is introduced. In addition, this paper proposes a cooperation-and-competition graph convolutional network (CC-GCN)-based model for predicting user behavior. Experiments demonstrate that the CC-GCN model not only improves the prediction accuracy of user behavior but also clearly reflects the game relationship between rumors and anti-rumors in the diffusion space.