【摘要】 <正> 一、前言聚类分析(Cluster Analysis)是数值分类(Numerical Taxoaomy)学中的一个新的分支,在很多领域得到了广泛应用,特别是系统聚类法(Hierachical Clustering Methods)是目前国内外使用最多的一种方法,类与类之间的距离的不同定义产生了系统聚类的不同方法,组平均(Group—avergae or average linking)法具有空间保持和单词性,在植物遗传育种中得到卓有成效的应用。自1936年,广义距离(Generalized Distance)提出以来,人们以广义距离作为分类统计量,用Rao(1952)的中枢压缩法(Pivotal Condensation)或用更多还原

【Abstract】 The reliability of fuzzy cluster analysis depents on the selection of lambda to a great extent. In the paper, based on 10 kinds of fuzzy cluster methods, classification was made using generalized distance discriminatory. The sum of deviation square within category and mutual information with geographical distribution for each classification were estimated. It showed that, while the sum of deviation square within category descended slowly as number of classification increasing, the value of lambda is at its best. Based on all mentioned above, the best division was made to 10 kinds of fuzzy cluster. It was got that Euclid’distance method is superior to the other.更多还原