不同系统聚类法在水化学分类中的应用

基于多元统计的数理方法——“聚类分析”为地下水化学分类的研究提供了合理、有效、快捷的手段。本文运用系统聚类法中最具代表性和典型性的最短距离法、重心法、类平均法以及离差平方和法这4种方法对重庆市八一隧道渗漏的19组水样进行了聚类分析，挖掘了数据间的内在联系，分析了水样间的相关关系，对比了各种方法的优缺点，探讨了不同方法的适用对象、应用条件和范围：经分析表明，最短距离法不适用于复杂的实际情况，重心法容易使聚类谱系图逆转，离差平方和法适用于样本数量较少的情况，而类平均法普遍适用。

Application of different hierarchical clustering methods to hydrochemical classification

As the traditional methods have their limitations and one-sided aspects, clustering analysis which based on multivariate statistical methods provides the rational, convenient and fast way to the groundwater hydrochemical classification study recently. The most representative and typical methods in hierarchical clustering analysis are the single linkage method, the eentroid method, the average linkage method and the Ward＇ s minimum-variance method. Based on the clustering analysis on the 19 groups of leakage water samples from Bayi Tunnel of Chongqing, the intrinsic relationships among the data and samples are excavated, the advantages and the disadvantages are compared and the applicable conditions and ranges of these 4 methods are analyzed. The results indicate that the single linkage method is not capable dealing with the complicated actual situations, the centroid method easily causes the reversal of the dendrogram, and the Ward＇ s method is applicable in the limited samples, while the average linkage method can competently be used in the general conditions.