摘要
提出了一种两阶段的聚类方法:Hybrid。第一阶段产生大小相同的圆形原子聚类;第二阶段合并原子聚类形成任意形状和大小的聚合聚类。在扩展边界时,不但考虑原子聚类间的距离,还考虑原子聚类的密度相似度。这样可以更好地排除“噪音”的影响,得到内部结构更加趋同的聚合聚类。
This paper presents a new clustering algorithm named Hybrid. Hybrid involves two phases: the first phase generates a set of round atom clusters with same size, and the second phase merges these atom clusters to create a set of molecule clusters with arbitrary size and shape. During the edge expanding process, Hybrid considers not only the distance between two atom clusters, but also the closeness of two atom clusters' densities. Therefore Hybrid can eliminate outlier effectively while maintaining more isomorphic molecule clusters.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2005年第13期1-3,50,共4页
Computer Engineering
基金
国家自然科学基金资助项目(60173058)
关键词
数据挖掘
聚类算法
原子聚类
聚合聚类
噪音
Data mining
Clustering algorithm
Atom cluster
Molecule cluster
Outlier
