D2RP-Net: dual-dynamics-based rumor propagation prediction network

In the era of digital connectivity, social media, as a powerful information transmission medium, weaves a complex web of propagation networks. This unprecedented speed of information flow, while facilitating efficient communication, also gives rise to challenges that are difficult to control, especially in the field of rumor propagation prediction, which has become a core topic in cybersecurity research. However, the open nature of social platforms and the cascading effect of information diffusion make it exceptionally difficult to accurately predict the propagation path of rumors and their potential impact range. To address the limitations of existing models in deeply understanding the interplay between local and global features in rumor propagation, which restricts prediction accuracy and comprehensiveness, we have innovatively proposed a Dual-Dynamics-based Rumor Propagation Prediction Network (D2RP-Net). This method cleverly integrates a temporal attention point process and dynamic equations to capture the microscopic propagation patterns in the temporal diffusion network; at the same time, it leverages the powerful capabilities of hypergraph convolutional networks to analyze the macroscopic propagation characteristics of the diffusion network. Furthermore, we have carefully designed a feature decoupling module aimed at effectively separating the macro and micro elements in the rumor propagation features, thereby enabling multi-scale and more refined rumor propagation prediction tasks.