Research on International Politics Combining Coverage Set Theory With Artificial Intelligence

Abstract

International conflicts in the post-Cold War era are characterized by multi-dimensional coupling and rapid evolution, and it is difficult for traditional models to capture their high-order nonlinear mechanisms. This paper proposes a dual-loop dynamic modeling framework that integrates Transformer and graph neural network (GNN) to collaboratively describe cross-domain events, trade dependence and multilateral voting relationships. Experiments show that the accuracy of this model is about 12% higher than that of baseline methods and can warn high-intensity conflicts three weeks in advance. Through interpretability analysis, three key driving factors of covenant chain strength, energy trade proportion, and negative emotional event density are identified, and the effectiveness of covering set theory in feature noise reduction and extraction is verified. This study provides quantitative tools for diplomatic risk early warning and significantly enhances the interpretability and credibility of artificial intelligence (AI) in international political analysis.