K-means聚类算法中聚类个数的方法研究  

The research method on the clustering number of K-means algorithm

作　　者：刘飞[1] 唐雅娟[1] 刘瑶[1] 

机构地区：[1]汕头大学电子工程系,广东汕头515063

出　　处：《电子设计工程》2017年第15期9-13,共5页Electronic Design Engineering

基　　金：国家自然科学基金面上项目(61471228);广东省重大科技计划项目(2015B020233018)

摘　　要：在数据挖掘算法中,K均值聚类算法是一种比较常见的无监督学习方法,簇间数据对象越相异,簇内数据对象越相似,说明该聚类效果越好。然而,簇个数的选取通常是由有经验的用户预先进行设定的参数。本文提出了一种能够自动确定聚类个数,采用SSE和簇的个数进行度量,提出了一种聚类个数自适应的聚类方法(简称:SKKM)。通过UCI数据和仿真数据对象的实验,对SKKM算法进行了验证,实验结果表明改进的算法可以快速的找到数据对象中聚类个数,提高了算法的性能。K-means clustering algorithm is a kind of common method of unsupervised learning in data mining. The more different objects are between clusters data, the more similar objects are within cluster data. The phenomenon shows that the cluster effect is better. The selection of cluster number is usually carried out by experienced users who set parameters. This dissertation proposed a SKKM method in which it determines the number of clusters through adopting SSE and the clustering number. The paper had a verification for SKKM algorithm by UCI data and lots of simulation experiments. The experimental result shows that the improved algorithm can determine the clustering number in the data objects, which achieved a noticeable gain of algorithmic performance.

关 键 词：K-MEANS算法 聚类个数 初始聚类中心 数据挖掘 K-means算法改进 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]