现代电子技术2024，Vol.47Issue(8)：1-8,8.DOI:10.16652/j.issn.1004-373x.2024.08.001
基于类簇合并的无参数密度峰值聚类算法
Nonparametric density peak clustering algorithm based on clusters merging
摘要
Abstract
The density peak clustering algorithm(DPC)is a simple and efficient clustering algorithm that can intuitively find the cluster centers by a decision graph and complete clustering.However,the cutoff distance and cluster centers of the DPC algorithm are both determined artificially and subject to significant subjective influence,resulting in uncertainty.To address the above issues,a nonparametric density peak clustering algorithm based on clusters merging(NDPCCM)is proposed.Based on the distribution characteristics of pairwise similarities among sample points,they are divided into intra-class similarity and inter-class similarity,and the cutoff similarity is determined automatically by means of intra-class similarity to avoid manually setting parameters.Based on the decreasing trend of the cluster center weights,the initial cluster centers are selected automatically to obtain the initial clusters.The initial clustering results are optimized by merging the initial clusters,which can improve the accuracy of clustering.The comparative experiments were conducted between the proposed algorithm and DPC,DBSCAN,and K-means algorithms on both artificial and UCI real datasets.The results show that the proposed algorithm does not require input parameters and can automatically obtain clusters,with better clustering performance than other algorithms.关键词
聚类分析/密度峰值聚类算法/初始类簇/类簇合并/相似度/聚类性能Key words
cluster analysis/density peak clustering algorithm/initial cluster/cluster merging/similarity/clustering performance分类
信息技术与安全科学引用本文复制引用
刘天娇,王胜景,袁永生..基于类簇合并的无参数密度峰值聚类算法[J].现代电子技术,2024,47(8):1-8,8.基金项目
国家自然科学基金资助项目(11201116) （11201116）
