现代信息科技2025，Vol.9Issue(13)：35-40,46,7.DOI:10.19850/j.cnki.2096-4706.2025.13.008
基于闭包感知的图神经网络链路预测方法
Link Prediction Method of Closure-Aware Graph Neural Network
摘要
Abstract
Message passing of Graph Neural Network(GNN)demonstrates powerful node representation capabilities in link prediction tasks,yet reducing interference from noisy edges in networks while improving link prediction performance remains a key issue.To solve this issue,this paper proposes a Closure-Aware Graph Neural Network link prediction method(Closure-Aware GNN).This method deeply explores the closure characteristics in networks,introduces the relationship between the shortest path between source and target nodes and their in-degrees and out-degrees,enhances the model's ability to identify real potential links,and suppresses interference from noisy edges simultaneously.Experimental results on 9 datasets show that the proposed method significantly improves the effectiveness and performance of link prediction.Evidently,closure plays an important role in enhancing link prediction accuracy and reducing the impact of noisy edges.关键词
链路预测/闭包感知/局部信息/图神经网络/复杂网络Key words
link prediction/Closure-Aware/local information/Graph Neural Network/complex network分类
信息技术与安全科学引用本文复制引用
凌非..基于闭包感知的图神经网络链路预测方法[J].现代信息科技,2025,9(13):35-40,46,7.基金项目
浙江经济职业技术学院应用技术研究所项目(X2023003) （X2023003）
