南京大学学报（自然科学版）2023，Vol.59Issue(6)：961-969,9.DOI:10.13232/j.cnki.jnju.2023.06.006
基于簇间连接的元聚类集成算法
Link based meta-clustering algorithm
摘要
Abstract
Clustering ensemble has become a popular research topic in data mining and machine learning.Despite significant progress in recent years,there are still two challenging problems in current researches on clustering ensemble.Firstly,most ensemble algorithms tend to study similarity at the object level,lacking the ability to mine cluster-level information.Secondly,many current ensemble algorithms focus only on the direct co-occurrence of objects within clusters,ignoring the relationship between clusters.To address these two problems,this paper proposes a link based meta-clustering algorithm(L-MCLA)which constructs a cluster similarity matrix based on Jaccard similarity,then refines this similarity matrix using connecting triples,and finally obtains the final result through graph partitioning and membership assignment.Theoretical analysis and experimental testing show that the proposed algorithm not only produces good clustering results,but also is less affected by the scale of clustering ensemble.关键词
簇间相似性/聚类集成/聚类/连接三元组/元聚类Key words
inter-cluster similarity/clustering ensemble/clustering/connecting triples/meta-clustering algorithm(MCLA)分类
信息技术与安全科学引用本文复制引用
杜淑颖,丁世飞,邵长龙..基于簇间连接的元聚类集成算法[J].南京大学学报（自然科学版）,2023,59(6):961-969,9.基金项目
国家自然科学基金(62276265,61976216),江苏高校"青蓝工程",江苏省高等职业院校专业带头人高端研修资助项目(2021GRFX074) （62276265,61976216）
