论文部分内容阅读
传统的系统聚类分析方法不能自动判断结果中适宜的类别数,本文应用两步聚类法与层次聚类法相结合的方式对25种汽车样本进行聚类分析,并利用类型内标准差与总体标准差的比较对聚类效果进行了检验。结果表明,使用这种组合聚类分析方式先科学、客观地确定聚类数目,再进行聚类分析,得到的三类数据样本集合,样本集合间各指标存在显著差异,所得到的结果与客观实际相符,证明了这种新聚类方式的可行性。
The traditional method of cluster analysis can not automatically determine the appropriate number of categories in the result. In this paper, 25 kinds of automobile samples are clustered by two-step clustering method and hierarchical clustering method. The standard deviation of the clustering results were tested. The results show that using this method of cluster analysis, we first determine the number of clusters scientifically and objectively, and then conduct cluster analysis. There are significant differences among the three types of data samples and the sample sets. The obtained results are objective and objective The actual match proves the feasibility of this new clustering method.