作者: F. -W. Gerstengarbe , P. C. Werner
DOI: 10.1007/BF00867981
关键词:
摘要: Cluster analysis contains several multivariate methods for the separation of patterns (clusters). The definition optimum or universally best cluster is an unresolved issue. Three are special importance: 1. statistical confidence separation. 2. optimal number clusters. 3. description internal structure. Two new addressing these problems presented. On basis nonhierarchical minimum-distance a method described that allows clusters in statistically well-founded way. This solves one and two. Using newly developed rank-sum analysis, solution to third problem possible. An example shows practicability proposed procedures.