不同相似度度量方式的随机数据聚类分析

Random data clustering analysis based on different similarity measures

采用经典的欧几里德距离、曼哈顿距离以及形状相似距离3种不同相似度度量方式,应用标准模糊C均值聚类算法在多个表示矩形对象的二维随机数据集上进行聚类,分析对比其相似度评估性能。聚类结果的分类统计表明,形状相似距离相比其他两种距离,能够考虑矩形对象的形状相似因素进行相似度评估。

In this paper, the classical Euclidean distance and Manhattan distance as well as the shape similarity dis- tance （SSD） are used as the similarity measure in the standard fuzzy c-mean clustering test on multiple rectangular ob- jects represented by two-dimension data. Clustering results are classified and statistical analyzed. Compared to the oth- er two kinds of distances, the shape similarity distance can consider the shape similar factors to assess similarity.