Abstract: Considering the disadvantage of dimension ruin and original clustering centers difficult to solve,this paper proposes an improving K-means clustering algorithm,which core is to improve traditional algorithm by lowering dimensions and searching based on density and diffused original clustering centers.The results of the experiment show that this algorithm can obtain better clustering result than traditional K-means clustering algorithm at clustering precision and stability.

Key words: K-means clustering algorithm, density, clustering centers

摘要： 针对传统的K-均值算法聚类时所面临的维数灾难、初始聚类中心点难以确定的缺点,提出一种改进的K-均值算法,其核心思想是通过降维、基于密度及散布的初始中心点搜索等方法改进K-均值算法。实验结果证明改进后的算法无论在聚类精度还是在稳定性方面,都明显优于标准的K-均值算法。

关键词: K-均值算法, 密度, 聚类中心