一种分层自适应快速K-means算法*

现今如何在大数据库中找到有用的数据类型已成为一个研究热点,而对数据库中分类簇的识别是该领域广泛研究的一个问题。提出一种分层自适应快速k-means(Hierarchical Adaptive Fast K-means,HAFKM)算法对图像数据库分类聚簇。HAFKM根据提出的分层策略构建一棵非平衡聚类树,通过自适应的方法CEC(Cluster Evaluation Criterion)确定了除根节点外的每棵子树的分支数目,而在聚类树的每一层聚类中使用一种提出的判别函数(the cost-function)在颜色直方图上根据颜色等级直接聚类,从而可以在整棵树上快速聚类。实验表明,HAFKM通过在非平衡树上逐层聚类,并且通过CEC准确判断聚类数目,可以快速、高效的实现数据库的分类聚簇。

A Hierarchical Adaptive Fast K-means Algorithm

Finding useful patterns in large datasets has become a research hotspot recently, and one of the most widely studied problems in this area is the identification of clusters in a dataset. We put forward the method of Hierarchical Adaptive Fast K-means(HAFKM) classification and clustering for image data base. HAFKM according to the proposed hierarchical strategy to build a unbalanced cluster tree, Through the method of adaptive CEC (Cluster Evaluation Criterion) determines every subtree branch number of the nodesexcept tne root node, and In each layer of the clustering tree using a proposed cost-function cluster on the color level histogram clustering directly, and then can fast cluster in the whole tree. Experiments show that HAFKM throuth the layered cluster in the unbalanced tree, and through CEC to determine the correct number of clusters, in the end can realize the classification of database fast and efficiently.