Abstract: As a new problem, data stream management has been focused on by a lot of scholars. Data clustering is an important field of it. Based on BIRCH cluster algorithm, this paper proposes a new algorithm M-BIRCH. It can overcome the problem of Birch that is not capable enough to cluster arbitrary shapes. In the experiment, it is shown that M-BIRCH is better than BIRCH in clustering quality.

摘要： 目前，数据流管理作为一种新兴课题正在逐渐受到国内外广大研究学者的重视，数据流聚类是其中的一个重要研究领域。本文基于BIRCH聚类算法，提出了一种M-BIRCH聚类算法，克服了BIRCH算法对非球形的聚类效果不佳等缺点。实验结果证明，M-BIRCH聚类算法在聚类质量上比BIRCH有较大提高。