作者: V. Estivill-Castro , M. E. Houle
DOI: 10.1007/S00453-001-0010-1
关键词: Data mining 、 Information extraction 、 Cluster (physics) 、 Cluster analysis 、 Combinatorial optimization 、 Exploratory data analysis 、 Delaunay triangulation 、 Data processing 、 Computer science 、 Medoid
摘要: In this paper we present a method for clustering geo-referenced data suitable applications in spatial mining, based on the medoid method. The is related to k -MEANS, with restriction that cluster representatives be chosen from among elements. Although general produces clusters of high quality, especially presence noise, it often criticized Ω(n 2 ) time requires. Our incorporates both proximity and density information achieve high-quality subquadratic time; does not require user specify number advance. bound achieved by means fast approximation objective function, using Delaunay triangulations store information.