Basic Consideration of Rough C-Medoids Clustering with Minkowski Distance
Basic Consideration of Rough C-Medoids Clustering with Minkowski Distance
复制标题
DOI：
10.1109/scisisis50064.2020.9322745
复制
发表时间：
2020-12
期刊：
影响因子：
--
通讯作者：
S. Ubukata;Atsushi Sugimoto;A. Notsu;Katsuhiro Honda
中科院分区：
文献类型：
--
作者：
S. Ubukata;Atsushi Sugimoto;A. Notsu;Katsuhiro Honda
聚类是自动分类和汇总数据的有用技术。特别地，基于粗糙集理论的聚类作为实现高度可靠的聚类的技术而受到关注。在本研究中，我们通过将粗糙集理论的视角引入到硬C-medoids（HCMdd；k-medoids）中，参考广义粗糙C-means（GRCM），提出粗糙C-medoids（RCMdd）聚类，来处理对象属于聚类的确定性、可能性和不确定性。此外，我们考虑使用 Minkowski 距离（欧几里德距离的推广）作为 RCMdd 中距离标准的基础。通过数值实验发现RCMdd可以像GRCM一样提取聚类的某些区域并进行可靠的聚类。通过引入中心点，RCMdd 实现了比 GRCM 更高的聚类性能。
Clustering is a useful technique for automatically classifying and summarizing data. In particular, clustering based on rough set theory is noticed as a technique for realizing highly reliable clustering. In this study, we propose rough C-medoids (RCMdd) clustering by introducing the perspective of rough set theory to hard C-medoids (HCMdd; k-medoids) referring to generalized rough C-means (GRCM), to deal with the certainty, possibility, and uncertainty of belonging of object to clusters. In addition, we consider using the Minkowski distance, which is a generalization of the Euclidean distance, as the base of distance criteria in RCMdd. Through numerical experiments, it was found that RCMdd can extract certain areas of clusters as well as GRCM and perform reliable clustering. By introducing the medoids, RCMdd achieved higher clustering performance than GRCM.