计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (22): 179-181.

• 数据库与信息处理 • 上一篇    下一篇

优化K-means初始聚类中心研究

毛韶阳1,2,李肯立2   

  1. 1.湖南人文科技学院 数学系,湖南 娄底 417000
    2.湖南大学 计算机与通信学院,长沙 410082
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-08-01 发布日期:2007-08-01
  • 通讯作者: 毛韶阳

Research of optimal K-means initial clustering center

MAO Shao-yang1,2,LI Ken-li2   

  1. 1.Department of Mathematics,Hunan Institute of Humanities,Science and Technology,Loudi,Hubei 417000,China
    2.School of Computer and Communication,Hunan University,Changsha 410082,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-08-01 Published:2007-08-01
  • Contact: MAO Shao-yang

摘要: K-means算法因为对初始中心依赖性而导致聚类结果可能陷入局部极小。基于密度的多中心聚类并结合小类合并运算的聚类算法解决了计算空间上的极小化,收敛进度上得到了控制,结果明显优于K-means的聚类结果。算法的每一次迭代都是倾向于发现超球面簇,尤其对于延伸状的不规则簇具有良好的聚类能力。

关键词: 聚类算法, K-means, 多中心聚类算法(MCA), 小类合并

Abstract: The K-means algorithm with the clustering dependence on the initial center may sink into the part smallest.The algorithm which using multi-centers clustering based on the density function and combining small cluster merger solves the computing space minimize,controls the rate of convergence.The experiment result surpass to K-means clustering result obviously.The every repeated operation incline toward discover super surface of sphere cluster.The algorithm has better clustering ability particularly to the irregular and extendable center.

Key words: clustering algorithm, K-means, Multi-seed Clustering Algorithm(MCA), merging small cluster