摘要
提出了一种改进的聚类分析算法,该算法采用类似中间聚类与最终聚类分布的思想,先对密集区域进行聚类,形成了K个聚类,然后再对相对分散的自由数据进行K-means聚类,使聚类分析在迭代过程中始终沿着最优的方向进行,减小了迭代次数,提高了收敛速度。该算法融合了网格聚类与K-均值聚类的优点,并且引入了一种新的划分网格的算法和新的计算密度阀值的函数。理论分析以及实验证明,改进算法的聚类过程达到了令人满意的效果。
An improved clustering analysis algorithm is proposed. Using the idea similar to half-finished clustering and final clustering distribution, the algorithm firstly clusters concentrated regions to get K clusters, and then clusters relatively scattered free data in K-means, which makes clustering analysis always follow optimal direction in iterative process, reduces iteration times and improves convergence speed. The algorithm integrates the advantages of grid-based clustering and K-means clustering, and introduces a new algorithm of partitioning grid and new function of computing density threshold. The theoretical analysis and experiments prove that the clustering process of the improved algorithm achieves satisfactory results.
出处
《计算机时代》
2010年第8期4-6,共3页
Computer Era
关键词
聚类分析
K-均值算法
网格聚类
融合聚类
clustering analysis
K-means algorithm
grid-based clustering
fusion clustering