利用降维排序求多维数据模糊聚类

摘要

本文根据数据间相似程度的定义，利用排序缩小范围，提出一种多维数据聚类的算法。它以一维数据的聚类方法为基础，通过对各映射子空间聚类相交，得到高雏聚类的近似解，进而求出精确解。该算法有较好的性能和较强的可伸缩性，且只需一个输入参数：置信水平λ，当置信水平λ变化时，可利用已有的计算结果。%According to the similitude definition between data, this paper proposes an efficient multi-dimension fuzzy clusteringalgorithm by dimension reducing and sorting. Starting with the clusters of data projected to one dimension ,the algorithm getsthe approximate multi-dimension fuzzy clusters through set intersection, and finally finds the exact multi-dimension fuzzy dus-ters. The scalable algorithm need only one input parameter: the credit levelλ, and when the λ is changed, it can uses the previ-ous results.