摘要
为了能够快速准确地发现自然分布的、任意形状密度变化的聚类,提出了基于近邻传播的快速扫描算法,该算法利用最近邻居关系的传递特性实现数据集合的完全聚类,简化了传统聚类方法的最近邻居判定和计算,优化了搜索过程,实现了快速聚类分析过程。通过与同类算法的比对验证,结果表明该算法对目标数据集合的任意分布特性有很好的适应能力。
In order to find all clusters which have the characteristics of natural distributions, arbitrary density and shape quickly and accurately, the paper present a new clustering algorithm, that is, the Fast Search Clustering Algo- rithm based on Affinity Propagation. Utilize the transmission characteristics among the nearest neighbors, the algorithm implement the full clustering on target data set. By simplify the computation and judge of the nearest neighbors among the traditional algorithms, and optimize the search process, realize the fast clustering. Compare experiments result with the other related works; find the new algorithm has the strong adaptability to the natural distribution data set.
出处
《华北电力大学学报（自然科学版）》
CAS
北大核心
2012年第5期93-96,共4页
Journal of North China Electric Power University：Natural Science Edition
基金
河北省社会科学基金资助项目(HB12YJ064)
关键词
近邻传播
自然分布
聚类分析
数据挖掘
affinity propagation
natural distribution
clustering analysis
data mining
