基于标准化变换的求和法:一种新的样品聚类分析方法

Summing method based on a standardized transformation:a new sample clustering methodology

本文目的是介绍一种新的样品聚类分析方法,即基于标准化变换的求和法,它也可归类于传统综合评价方法体系之中。具体地说,第1步,对全部定量变量进行同趋势化变换;第2步,选择一种合适的定量变量标准化变换方法,使第3步中求得的"综合指标"的标准差达到最大值;第3步,求每个样品按综合指标上的数值,即求每个样品在全部标准化变换后的定量指标上取值之和;第4步,将全部样品按综合指标上的取值由大到小(整体为高优指标)或由小到大(整体为低优指标)排序,此顺序就是由优到劣的顺序。基于统计学基础知识和任何两个样品之间易于判定优劣的基本常识,本文提出了评价样品优劣排序合理性的两条标准。以此标准为依据,得出本文方法比其他七种综合评价方法(本质上是样品排序或样品聚类)的排序结果都更为合理。

This article aims to introduce a new method for the sample clustering analysis,entitled the summing method based on a standardized transformation, which can also be categorized into the traditional method system of the generalizing comprehensive evaluation. Briefly speaking as follows step 1, all the quantitative variables are transformed with the same trend ; step 2 , select a suitable quantitative variable normalization method, so that the standard deviation of the ＂ comprehensive index＂ to reach the maximum in the next step; step 3, find the value of each sample on the comprehensive index, that i s,find the summation of the standardized values of the quantitative indices of all the samples ; step 4 ,sequence all the samples in accordance with the numeric size of the comprehensive indicators from large to small （the entire indicators considered to be high priority index） or from small to large （ the entire indicators considered to be low priority index） , and this sequence indicates the order from superiority to the inferiority. On the basis of the statistical knowledge and the common knowledge of pointing the bigger one between 2 values, this article raises 2 criteria for evaluating the sequence rules. Based on this standard, we conclude that this new method is superior to the other 7 methods of the comprehensive evaluation which are essentially for sample sorting or sample clustering.