电子科技2025，Vol.38Issue(11)：34-41,8.DOI:10.16180/j.cnki.issn1007-7820.2025.11.005
显著子图挖掘方法
Research on Significant Subgraph Mining Method
摘要
Abstract
In view of the problem that subgraph neural networks ignore important structural semantics in subgraph sampling,which leads to low accuracy of graph classification tasks and lack of generalization of subgraph neural net-works,a significant subgraph mining method is proposed in this study,aiming to improve the accuracy and generaliza-tion performance of subgraph neural networks by adaptive mining important subgraph structural semantics.The intimacy sampling module based on PageRank algorithm is used to calculate the intimacy of nodes to self-adaptively quantify the importance of nodes,construct significant subgraphs according to the importance of nodes,and select significant sub-graphs according to the intimacy between subgraphs.The significant subgraph composed of important nodes contains im-portant structural information in the original graph.By learning the significant subgraph,the subgraph neural network can obtain the original graph representation more comprehensively,thus improving the accuracy and generalization of the subgraph neural network in downstream tasks.The proposed method achieves good experimental results in all five data sets,and the accuracy rate in MUTAG data set reaches 96.80%.关键词
人工智能/图神经网络/子图挖掘/图分类/亲密度采样/子图亲密度/自适应挖掘/显著子图Key words
artificial intelligence/graph neural networks/subgraph mining/graph classification/affinity sam-pling/subgraph affinity/adaptive mining/significant subgraph分类
计算机与自动化引用本文复制引用
刘澳龙,唐向红,陆见光..显著子图挖掘方法[J].电子科技,2025,38(11):34-41,8.基金项目
贵州省科学技术基金(QKHJC-ZK[2021]YB271,QKHJC-ZK[2021]YB015) （QKHJC-ZK[2021]YB271,QKHJC-ZK[2021]YB015）
贵州省科技支撑项目(QKHZC[2022]YB074) The Science and Technology Foundation of Guizhou(QKHJC-ZK[2021]YB271,QKHJC-ZK[2021]YB015) （QKHZC[2022]YB074）
Guizhou Science and Technolo-gy Support Project(QKHZC[2022]YB074) （QKHZC[2022]YB074）
