节点文献
有关距离判别理论方法的补充和讨论
A Complement and Some Discussions to the Theory of Distance Discrimination
【摘要】 距离判别理论中,通常采用重心距离来定义类与类之间的距离对待判样品进行判别。对新样品实行判别,将其归入系统聚类形成的分类,如果仍采用重心距离判别法,会由于没有与原有聚类时所用的类与类之间的距离相一致而产生误判。提出对基于系统聚类分类结果的距离判别理论方法的补充,把系统聚类中的八种类与类之间距离的概念引入到距离判别方法中。从而使距离判别中类与类距离的定义与系统聚类中相一致,通过实例分析,证明增强了距离判别的可靠性。
【Abstract】 In the theory of distance discrimination,we often do discrimination analysis by centroid distance method.If we do discrimination analysis on the basis of hierarchical cluster’s results,one can make mistakes due to misuse of distances in cluster analysis which is not consistent with the distance in hierarchical cluster.I suggest a complement to the theory of distance discrimination-based on hierarchical cluster’s results: Doing discrimination analysis also use all eight cluster’s distance in hierarchical cluster analysis.Thus we can use the same distance among clusters in the both methods,I have proved the enhancing of the reliability in distance discrimination analysis by taking an example.
【Key words】 distance discrimination; centroid distance; hierarchical cluster;
- 【文献出处】 统计与信息论坛 ,Statistics & Information Forum , 编辑部邮箱 ,2007年04期
- 【分类号】F224
- 【被引频次】9
- 【下载频次】224