密度峰值聚类算法研究综述  

Survey of Density Peak Clustering Algorithm

作　　者：王森 邢帅杰 刘琛 Wang Sen;Xing Shuaijie;Liu Chen(School of Science,East China Jiaotong University,Nanchang 330013,China)

机构地区：[1]华东交通大学理学院,江西南昌330013

出　　处：《华东交通大学学报》2023年第1期106-116,共11页Journal of East China Jiaotong University

基　　金：江西省自然科学基金项目(20192ACBL20010)。

摘　　要：密度峰值聚类(DPC)是一种新提出的基于密度和距离的聚类算法,由于其原理简单,无需迭代和能处理形状数据集等优点,正在数据挖掘领域得到广泛应用。但DPC算法也有着一定的缺陷,如:对截断距离参数敏感,初始聚类中心的选择非自动化,后续标签分配存在链式问题,时间复杂度较高等。文章对DPC算法的研究现状进行了总结与整理,首先介绍了DPC的算法原理和流程;其次,针对DPC算法的不足对DPC算法的优化进行概括和分析,指出了优化算法的核心技术以及优缺点;最后,对DPC算法未来可能面对的挑战和发展趋势进行展望。Density peak clustering(DPC)is a novel clustering algorithm based on density and distance.It is widely used in the field of data mining because of its simple principle,no iteration and the ability to process shape datasets.However,DPC algorithm also has some defects,including the sensitive cutoff distance parameter,non-automatic selection of initial clustering center,the chain problem in subsequent allocation and high time complexity.This paper summarizes the research status of DPC algorithm.Firstly,it introduces the principle and process of DPC algorithm.Secondly,in view of the deficiencies of DPC algorithm,the optimization of DPC algorithm is summarized and analyzed,and the core technology,advantages and disadvantages of the optimization algorithm are pointed out.Finally,the possible challenges and development trend of DPC algorithm in the future are concluded.

关 键 词：聚类算法 密度峰值 截断距离 初始聚类中心 微簇合并 时间复杂度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]