Construction of Event Graph of Floods Based on Graph Attention Network
Abstract
Emergencies are characterized by suddenness, destructiveness, urgency and uncertainty, which will bring immeasurable negative impact to society. Flood is a typical emergency event. This paper constructs an event graph of floods based on the graph attention network by Henan flood 2021 and analyzes the key events of flood and the theme of flood rescue. This study innovatively applies the graph attention network GAT to flood event modeling, fills the research gap of event graph in flood disaster field, and provides actionable technical support for flood emergency decision-making. Experimental results show that our GAT-based model achieves an accuracy of 0.894 in event relation extraction, outperforming the baseline by 3.1%.
Fields of Study & Subjects
Topic Modeling
Flood Risk Assessment and Management
Advanced Graph Neural Networks
Publication details
- Journal / source
- Frontiers in artificial intelligence and applications
- Publication date
- 27 January 2026
- Document type
- Book chapter