一种基于稀疏搜索的激活噪声快速聚类算法  

An activating noise fast clustering via sparse search

作　　者：郭林亮 韩旭明 张逸航 GUO Lin-liang;HAN Xu-ming;ZHANG Yi-hang(School of Mathematics and Statistics,Changchun University of Technology,Changchun 130012,China;College of Information Science and Technology,Jinan University,Guangzhou 510632,China;College of Computer Science and Engineering,Changchun University of Technology,Changchun 130012,China)

机构地区：[1]长春工业大学数学与统计学院,吉林长春130012 [2]暨南大学信息科学与技术学院,广东广州510632 [3]长春工业大学计算机科学与工程学院,吉林长春130012

出　　处：《东北师大学报（自然科学版）》2022年第3期55-59,共5页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61472049);吉林省科技发展计划技术攻关项目(20190302071GX);吉林省产业技术研究与开发专项项目(2019C053-11)。

摘　　要：针对如何分配噪声和近邻类连接处的数据点影响聚类结果的问题,提出了一种基于稀疏搜索的激活噪声快速聚类算法(ANSC).ANSC能够激活数据中的噪声,以步长方式快速稀疏搜索互连点来共建子簇,并且只需要设置近邻个数.ANSC可以根据噪声的分布特点来判断其类型.在人工和真实数据集上进行实验,结果表明,所提出的方法在有效性和效率方面优于其他的聚类算法.How to deal with noise and joint points between clusters is the key factor affecting clustering results.To address the problem, an activating noise fast clustering via sparse search(ANSC) is proposed in this paper.ANSC can activate noise in the data and construct sub-clusters by fast sparse search co-connection points in step size, and only need unique number of neighbors.Moreover, ANSC is able to detect the type of noise according to the distribution characteristics of noise.Experimental results on synthetic and real world datasets show that the proposed method is superior to other clustering algorithms in effectiveness and efficiency.

关 键 词：聚类 噪声 步长 稀疏搜索 互连点 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]