基于闭包的聚类判别方法研究  

Clustering discrimination based on core-sets

作　　者：彭晓琳[1] 黄樟灿[1] 朱洁[1] 

机构地区：[1]武汉理工大学理学院,武汉430070

出　　处：《计算机工程与应用》2009年第18期191-193,共3页Computer Engineering and Applications

摘　　要：区别于传统的聚类方法,提出了以类为起点,通过构造闭包进行聚类的新方法,并建立了聚类判别模型,此模型给出了对于闭包间的交叉区域的检验点的判别准则。然后针对二维的聚类问题,提出了以最小圆为闭包的聚类判别模型,并对乳房肿瘤病例进行数值实验。对于乳房肿瘤病例,首先进行了指标选取、数据预处理,然后以最小圆为闭包建立了模型,最后对69个待检测数据进行检验,结果误判率为4.35%。This paper presents one clustering diserimination method which is different from the previous methods.It starts with the categories,constructs core-sets for clustering discrimination,and establishes a model to it.This paper uses the smallest circle as the core-sets for clustering discrimination in the 2-dimensional space and does numerical experiment on the diagnosis of breast cancer.In this model,this paper first selects the indicators,processes the data,and then takes the smallest circle as coresets to establish the model, at last tests another 69 data.The rate of the misjudgment is 4.35%.

关 键 词：闭包 聚类中心 聚类判别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]