Zhang Mingrui, Zhang Wei, Sicotte Hugues, Yang Ping
Computer Science Department, Winona State University, MN 55987, USA.
Annu Int Conf IEEE Eng Med Biol Soc. 2009;2009:3865-8. doi: 10.1109/IEMBS.2009.5332582.
One of the major challenges in unsupervised clustering is the lack of consistent means for assessing the quality of clusters. In this paper, we evaluate several validity measures in fuzzy clustering and develop a new measure for a fuzzy c-means algorithm which uses a Pearson correlation in its distance metrics. The measure is designed with within-cluster sum of square, and makes use of fuzzy memberships. In comparing to the existing fuzzy partition coefficient and a fuzzy validity index, this new measure performs consistently across six microarray datasets. The newly developed measure could be used to assess the validity of fuzzy clusters produced by a correlation-based fuzzy c-means clustering algorithm.
无监督聚类的主要挑战之一是缺乏评估聚类质量的一致方法。在本文中，我们评估了模糊聚类中的几种有效性度量，并为模糊 c 均值算法开发了一种新的度量，该算法在其距离度量中使用皮尔逊相关性。该度量是基于簇内平方和设计的，并利用了模糊隶属度。与现有的模糊划分系数和模糊有效性指标相比，这种新度量在六个微阵列数据集上表现一致。新开发的度量可用于评估基于相关性的模糊 c 均值聚类算法产生的模糊聚类的有效性。
