基于加权相似性的BIRCH聚类算法  

BIRCH Clustering Algorithm Based on the Weighted Similarity

作　　者：邹杰涛[1] 赵方霞[1] 汪海燕[1] 

机构地区：[1]北方工业大学理学院,北京100144

出　　处：《数学的实践与认识》2011年第16期118-124,共7页Mathematics in Practice and Theory

摘　　要：BIRCH方法是一个集成的层次聚类方法.它克服了凝聚层次聚类方法所面临的两个难点:可伸缩性和不能撤销前一步工作的问题.基于BIRCH聚类的多阶段聚类算法思想,结合基于权重的欧式距离度量和基于划分的K-means算法,提出了一种基于加权相似性的BIRCH聚类方法,并将方法应用在时间序列的气象数据分析中.BIRCH is an integrated method of hierarchical clustering. It conquers two difficulties of condensation hierarchical clustering, which are the problems of scalability and rollback. Based on the idea of multi-stage clustering of the BIRCH algorithm, with the weight-based Euclidian distance metric and partition-based K-means algorithm, this paper proposes the BRICH clustering algorithm based on the weighted similarity, and the method is applied into the time series analysis of meteorology data.

关 键 词：数据挖掘 聚类分析 气象数据 时间序列 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]