聚类金字塔树:一种新的高维空间数据索引方法  

Clustering Pyramid-Tree: A New Index Structure for High-Dimensional Data Spaces

作　　者：张海勤[1] 欧阳为民[2] 蔡庆生[1] 

机构地区：[1]中国科学技术大学计算机科学技术系,合肥230027 [2]安徽大学计算中心,合肥230039

出　　处：《中国科学技术大学学报》2001年第6期707-713,共7页JUSTC

基　　金：国家自然科学基金资助项目 (6 0 0 75 0 15 )

摘　　要：提出了一种新的有效的高维空间数据索引方法———聚类金字塔树 .它先对不均匀分布数据进行聚类处理 ,然后对聚类的结果实施金字塔分割和存储 ,由此建立一种有效的索引结构 .文中给出了聚类金字塔树的几种查询算法 .实验证明 :处理不均匀分布数据时 ,聚类金字塔树无论在页面访问次数 。A new efficient index structure is presented for high-dimensional data space-Clustering Pyramid-tree. Due to the low efficiency of Pyramid-tree in processing asymmetric distributed data, especially clustered data, Clustering Pyramid-tree is proposed as an extension of the Pyramid-tree. To the asymmetric distributed data, Clustering Pyramid-tree technique first applies clustering algorithms and indexes the clustered results, then it makes pyramid partitioning for every cluster and creates an efficient index structure for the partitions. Clustering Pyramid-tree technique uses several index trees instead of one tree of Pyramid-tree technique. Several corresponding query algorithms, including range query and nearest neighbor query, are presented in detail. The experiments show that the Clustering Pyramid-tree technique outperforms the Pyramid-tree technique both in page accesses and CPU total time when processing asymmetric distributed data.

关 键 词：高维数据空间 不均匀数据分布 聚类金字塔树 范围查询 高维空间索引法 

分 类 号：TP311.132[自动化与计算机技术—计算机软件与理论]