摘要
针对传统基于距离的聚类算法所存在的缺点,将万有引力和牛顿第二运动定律思想引入到聚类过程中,提出了一种改进的基于密度的引力聚类算法GCABD.该算法可以自动决定目标数据集中的簇的个数,并且能发现任意形状的簇且可以过滤"噪声"数据.实验结果表明,所提出的GCABD算法的聚类效果和精度均比典型的K-means算法好,提高了聚类质量.
Directing against the drawbacks of traditional algorithm based on distance,the paper introduces gravitation and Newton second law of motion into the process of clustering, and proposes an improved algorithm GCABD （Gravitational Clustering Algorithm Based on Density）. This algorithm can decide automatically the number of clusters in the target data set, and find any clusters with arbitrary forms and filter the noisy data. The experimental results show that GCABD algorithm is superior than typical K-means algorithm in clustering effect and precision, enhances the clustering quality greatly.
出处
《河南科学》
2008年第11期1400-1404,共5页
Henan Science
关键词
数据挖掘
聚类分析
聚类算法
引力
data mining
clustering analysis
clustering algorithm
gravitation
