An information propagation model based on rumor-debunking and multi-dimensional influence

To address the challenges of data sparsity and the complex interplay of rumor and debunking information in the early stages of rumor propagation, an Information Propagation Model Based on Rumor-Debunking and Multi-Dimensional Influence is proposed. Firstly, to mitigate data sparsity, a three-dimensional tensor model of “active user-potential user-interaction behavior” is constructed. The three-dimensional tensor model leverages the low-rank approximation properties of tensor decomposition to extract hidden relationships and weak correlation patterns between potential and active users, effectively alleviating the issue of sparse rumor data. Secondly, game theory is introduced to quantify the influence of multi-level interactions. A quantification framework is designed for multi-type topic message drivers. This framework considers the competitive relationships among different messages. A dynamic multi-message-oriented interaction strategy is employed to model the influence of multi-source information flow decisions quantitatively. Finally, considering multiple influencing factors in rumor dissemination and the dynamic nature of network topology, a multi-dimensional influence coupling matrix is constructed. This matrix is used to reconstruct the user relationship networks. Additionally, a time-aware graph attention mechanism is introduced to model the sequential dependencies of user interactions. This mechanism leverages spatiotemporal joint embedding to enhance the ability to capture the dynamic evolution of user decision-making. Experimental results demonstrate that the proposed model effectively captures the influence of multi-dimensional factors on user behavior during rumor-debunking topic propagation and significantly enhances the prediction of topic dissemination dynamics. Compared to baseline models, the proposed method improves prediction accuracy by an average of 4.8%.