Article: A spatio-temporal concept-based knowledge graph model for inferring causal relationships in historical events Journal: International Journal of Reasoning-based Intelligent Systems (IJRIS) 2026 Vol.18 No.16 pp.31 - 50 Abstract: This paper proposes a causal-aware knowledge graph reasoning model that integrates temporal concepts, aiming to address the challenge in historical event analysis where traditional methods neglect the coupling of time and space, resulting in biased causal inference. This model incorporates a time decay function and a spatial graph attention mechanism to embed the 'long-term' historical perspective and geographical proximity into event encoding, and combines Granger causality testing to construct candidate causal subgraphs. On public datasets, the causal reasoning area under the curve of the model in this paper reaches 0.925 and 0.908 respectively, which is 5.2% and 4.8% higher than the baseline model historical temporal causal graph attention network; normalised discounted cumulative gain @10 reaches 0.87 and 0.83, outperforming several baseline temporal knowledge graph methods in the experimental settings. The study reveals the crucial role of temporal concepts in stripping confounding factors and reconstructing historical causal chains, providing interpretable analytical tools for computational history. Inderscience Publishers - linking academia, business and industry through research

Title: A spatio-temporal concept-based knowledge graph model for inferring causal relationships in historical events

Authors: Xiaofeng Ren

Addresses: College of Political Science Law and History, Baoji University of Arts and Sciences, Baoji, 721013, China

Abstract: This paper proposes a causal-aware knowledge graph reasoning model that integrates temporal concepts, aiming to address the challenge in historical event analysis where traditional methods neglect the coupling of time and space, resulting in biased causal inference. This model incorporates a time decay function and a spatial graph attention mechanism to embed the 'long-term' historical perspective and geographical proximity into event encoding, and combines Granger causality testing to construct candidate causal subgraphs. On public datasets, the causal reasoning area under the curve of the model in this paper reaches 0.925 and 0.908 respectively, which is 5.2% and 4.8% higher than the baseline model historical temporal causal graph attention network; normalised discounted cumulative gain @10 reaches 0.87 and 0.83, outperforming several baseline temporal knowledge graph methods in the experimental settings. The study reveals the crucial role of temporal concepts in stripping confounding factors and reconstructing historical causal chains, providing interpretable analytical tools for computational history.

Keywords: space-time knowledge graph; causal reasoning; historical event analysis; graph neural network.

DOI: 10.1504/IJRIS.2026.154387

International Journal of Reasoning-based Intelligent Systems, 2026 Vol.18 No.16, pp.31 - 50

Received: 06 Mar 2026
Accepted: 09 Apr 2026
Published online: 25 Jun 2026 *

Title: A spatio-temporal concept-based knowledge graph model for inferring causal relationships in historical events

Keep up-to-date