论文部分内容阅读
引入序关系保持的思想,即层次聚类的簇间距离度量应该能够最大限度地维护样本点间的原始距离排序关系。定义了样本点对序关系的概念和序关系损失度量,证明了序关系损失度量可用做聚类的目标准则函数和聚类结果质量的评价标准。利用序关系损失的概念扩展出两种簇间距离度量,实现了基于序关系保持的层次聚类算法(order-preserving based hierarchical clustering algorithm,OPHCLUS)。实验仿真证明了OPHCLUS对聚类质量提升的有效性。
The introduction of the idea of ordinal relationship preservation, that is, the hierarchical clustering distance measure between clusters should be able to maximize the maintenance of the original distance between the sample points Sort relationship. We define the concept of sample point order relation and the loss of ordinal relations, and prove that the ordinal relation loss measure can be used as clustering criteria and evaluation criteria of clustering quality. Based on the concept of order relation loss, this paper extends the distance measure between two clusters and realizes the order-preserving based hierarchical clustering algorithm (OPHCLUS). Experimental simulation proves the effectiveness of OPHCLUS in improving clustering quality.