摘要
针对k-means算法的缺陷,提出了一种新的多中心聚类算法。运用两阶段最大最小距离法搜索出最佳初始聚类中心,将原始数据集分割成小类后用合并算法形成最终类,即用多个聚类中心联合代表一个延伸状或者较大形状的簇。仿真实验表明:该算法能够智能地确定初始聚类种子个数,对不规则状数据集进行有效聚类,聚类性能显著优于k-means算法。
A novel multiseed clustering algorithm was proposed aiming at shortcomings of k-means algorithm. This algorithm could find optimal initial starting points applying iterative max-rain distance means and then combined the small clusters from given data set into final ones, for an elongated or large cluster could be considered as the union of a few small distinct hyperspherieal clusters. Experimcntal results demonstrate that the improved algorithm can automatically obtain the number of initial clusters, be effective on data set of irregular shapes and lead to better solutions than k-means algorithm.
出处
《计算机应用》
CSCD
北大核心
2006年第6期1425-1427,共3页
journal of Computer Applications
关键词
聚类
最大最小距离法
多中心
抽样
clustering
max-min distance means
multiseed
sampling