动态聚类的一种新方法——最小组内平方和法

A New Dynamic Clustering Method-MinSSw Method

本文提出一种以组内平方和之和最小为标准的新的动态聚类方法,简称为最小组内平方和法(MinSS_w法).Monte-Carlo试验表明,该法能有效地调整初始分组中的个体.使之达到合理的分类,并有着良好的稳健性;而目前常用的重心法(K-means法)的动态聚类结果,却是相当不稳定的.上述结论也为一个含14个小麦品种的聚类实例所证实.

A new dynamic clustering method which minimizes the total sum of squares within groups, the MinSS_W method, along with its algorithm was developed. The Monte-Carlo sampling was used to investigate the properties of the new method and to make some comparisons with K-means method. The new method can adjust entries in ini- tial groups into proper placement effectively. The performance of the new method is much more robust than that or K-means method in all situations. This is also demonstrated by a real example of partitioning 14 wheat varietics into 5 groups.