基于Leader的K均值改进算法  

Improved K-means clustering algorithm based the Leader algorithm

作　　者：张琼[1] 张莹[1] 白清源[1] 谢丽聪[1] 谢伙生[1] 

机构地区：[1]福州大学数学与计算机科学学院,福建福州350002

出　　处：《福州大学学报（自然科学版）》2008年第4期493-496,共4页Journal of Fuzhou University(Natural Science Edition)

基　　金：福建省教育厅科研资助项目(JB07022;JB06023);福州大学科技发展基金资助项目(2006-XQ-22;XRC-0511)

摘　　要：研究了K均值算法中初始聚类中心的选择对算法本身聚类精度及效率的影响,并提出了改进的算法(LK算法,Leader+K-means).LK算法中的初始聚类中心选择不是随机的,而是利用Leader算法得到若干个初始类中心,然后选择包含数据项最多的k个类中心,作为K均值算法的初始类中心.实验结果表明,LK算法在聚类结果的稳定性和正确率方面都是有效可行的.By researching in the relations between the initial means of clusters and the efficiency of clustering, the improved K - means clustering algorithm （ the LK algorithm, Leader ＋ K - means） is proposed. The LK algorithm is better since the initial means is not random selected. At first, it gains several initial means by means of the Leader algorithm, and then selects the k means containing the most data items regarded as the initial means. According to the experiment, the improved K -means clustering algorithm can get higher stability and accuracy .

关 键 词：聚类 K均值算法 Leader算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]