Hybrid knowledge reasoning over knowledge hypergraph:Inductive,deductive,and abductiveOA
Traditional knowledge reasoning methods,which are predominantly reliant on static rules and structured data,often struggle to adapt to the ambiguity and dynamic evolution of real-world scenarios.To overcome these limitations,this study proposes a novel reasoning framework based on a three-layered knowledge hypergraph.Core innovation lies in the synergy of inductive,deductive,and abductive reasoning mechanisms to enhance both reliability and interpretability.Specifically,hypergraph-based inductive reasoning extracts robust evolutionary patterns by mining the historical subgraph structures.Deductive reasoning ensures transparency by constructing tree-shaped inference paths,whereas abductive reasoning establishes causal traceability by forming evidence chains from historical contexts.Experimental evaluations on the Integrated Crisis Early Warning System(ICEWS)dataset demonstrate that the proposed approach significantly outperforms existing methods in terms of accuracy and interpretability,thereby offering a scalable solution for complex event analysis.
Ling Tian;Lei Gao;Ben Zhang;Xiao Liu;Yi-Nong Shi;Hui Gao
School of Computer Science and Engineering,University of Electronic Science and Technology of China,Chengdu,611731,ChinaSchool of Computer Science and Engineering,University of Electronic Science and Technology of China,Chengdu,611731,ChinaSchool of Computer Science and Engineering,University of Electronic Science and Technology of China,Chengdu,611731,ChinaSchool of Computer Science and Engineering,University of Electronic Science and Technology of China,Chengdu,611731,ChinaSchool of Computer Science and Engineering,University of Electronic Science and Technology of China,Chengdu,611731,ChinaSchool of Computer Science and Engineering,University of Electronic Science and Technology of China,Chengdu,611731,China
信息技术与安全科学
Abductive reasoningDeductive reasoningInductive reasoningKnowledge hypergraphsKnowledge reasoning
《Journal of Electronic Science and Technology》 2026 (2)
P.11-26,16
supported by the National Natural Science Foundation of China under Grant No.62376055.
评论