基于PCA的H-K聚类算法研究  

Study on PCA based Hierarchical k-means Clustering Algorithm

作　　者：何莹[1] 秦亮曦[1] 

机构地区：[1]广西大学计算机与电子信息学院,南宁530004

出　　处：《微计算机信息》2012年第6期163-165,共3页Control & Automation

摘　　要：传统的H-K聚类算法将层次聚类算法和k-means聚类算法有机结合起来,从而使得H-K聚类算法具有单个聚类算法所不具有的诸多优点。为了将H-K聚类算法更好地应用于对高维数据集的聚类中,以缓解维度灾难问题,本文应用PCA(主成分分析)方法对H-K算法进行改进,提出新的聚类算法PCAHK。该算法首先采用PCA方法,将高维数据投影到较低维空间中,再对降维后的数据进行H-K聚类。实验表明,在对高维数据集进行聚类时,与传统的H-K算法相比,PCAHK算法的性能明显提高。The traditional H-K clustering algorithms combined the hierarchical clustering algorithm and the k-means clustering algorithm, made H-K clustering algorithm have many merits which a sirigle clustering algorithm doesn＇t achieve. In order to make H-K clustering algorithms applied better in the Clustering of high-dimensional data sets and alleviate dimension disaster problem, in this paper, it is applied PCA method to H-K clustering algorithm for improvement, and proposed a new algorithm which we named as PCAHK. The algorithm firstly adopts PCA method to project high dimensional data into the lower dimensional space, then executes H-K clustering on the low dimensional data sets. The experiments show that the PCAHK algorithmget better performance than the traditional H-K clustering algorithm for the clustering of high-dimensional data sets.

关 键 词：聚类 高维数据集 PCAHK 轮廓系数 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]