Landslide Knowledge Representation Based on Hypergraph Theory

Abstract

High‐precision, multi‐source landslide monitoring data are crucial for disaster prevention and management. These datasets provide explicit landslide descriptions. Integrating this data with landslide mechanisms, using knowledge graphs can enhance early monitoring and forecasting. However, traditional knowledge graphs often oversimplify landslide knowledge, failing to capture the complexity of geological environments and landslide evolution. Spatio‐temporal knowledge graphs face challenges in representing intricate relationships. A hypergraph (HG), where an edge connects multiple nodes, offers a better representation of these complexities. This paper proposes an HG‐based method for landslide knowledge representation, organizing multi‐source information and knowledge through binary or multiple relationships under specific temporal and spatial conditions. A case study of the Miaodian landslide, which experienced multiple sliding events, shows that the proposed landslide knowledge HG outperforms other knowledge graphs like YAGO, Geographic Knowledge Graph (GeoKG), and Geographic Evolutionary Knowledge Graph (GEKG) in completeness, accuracy, and redundancy, demonstrating its effectiveness.