任意形状聚类的SPK-means算法  

SPK-means:a clustering algorithm for arbitrary shapes

作　　者：侯延琛 赵金东 HOU Yanchen;ZHAO Jindong(School of Computer and Control Engineering,Yantai University,Yantai 264010,Shandong,China)

机构地区：[1]烟台大学计算机与控制工程学院,山东烟台264010

出　　处：《山东大学学报（工学版）》2023年第2期87-92,101,共7页Journal of Shandong University（Engineering Science）

基　　金：国家自然科学基金项目(61972360);山东省自然科学基金(ZR2020MF148,ZR2020QF108)。

摘　　要：针对K-means聚类算法仅以质心作为聚类依据,在处理非圆球形数据集时效果不理想,数据集的形状特性未得到体现的问题,提出一种基于形状的K-means算法(shape K-means,SPK-means)。将判定点到不同簇中质心以及点到不同簇的最近边缘点的距离作为判定规则,使其具备对任意形状的数据集进行聚类的功能。设置两种不同的数据集进行聚类试验,结果表明,SPK-means聚类算法在处理非圆球形数据集时,其结果遵循原数据集的形状特征。Considered that the Kmeans clustering algorithm only took the center of mass as the clustering basis,the effect was not ideal when processing nonspherical data sets,and the shape characteristics of the data sets were not reflected.This research proposed an improved Kmeans algorithm(shape Kmeans,SPKmeans)which took the distance between the decision point and the center of mass in different clusters and the nearest edge point from the point to different clusters as the decision rule.This enabled it to cluster data sets of arbitrary shapes.Two different data sets were set for clustering in the experiment,and the results showed that the results of SPKmeans clustering algorithm followed the shape characteristics of the original data set when processing noncircular spherical data sets.

关 键 词：K-MEANS 高精度 边缘点 形状 快速聚类 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]