Key Node Identification in Dynamic Heterogeneous Hypergraphs for Sudden-Onset Disaster Public Opinion

In sudden disaster events, the dissemination of social media information has distinct stages. Identifying the key dissemination nodes at each stage is of great significance for tracing the source of the information and implementing emergency interventions. This paper divides the event evolution process into 72-hour time windows and constructs three types of nodes (users, content, and topics) and two types of hyperedges (content resonance circles and topic participation circles) in a dynamic heterogeneous hypergraph. It proposes a three-component influence score IS composed of content participation degree, topic participation intensity, and structural hole bridging degree. The weights of each component are adaptively determined based on the correlation between each component and the communication ability. Taking the Weibo data of 2023 Zhuozhou Rainstorm as an example, the propagation simulation verification is carried out using the independent cascading model. The results show that IS is superior to degree centrality, k-core, HITS and other structural baselines in all 7 snapshots, leading by 12.8 points relative to the strongest baseline and statistically significant; its identification gain is concentrated in medium and low-level users, indicating that this method can discover efficient disseminators that are easily overlooked by the structural baselines.