基于聚类分解的高维度量空间索引B+-Tree.pdf

ISSN RUXUEW 1000—9825，CODEN Journal ofSoftware，V01．19，No．6，June2008，PP．1401-1412 hnp：／／www．jos．org．cn DOI：10．3724／SP．J．1001．2008．0140l @2008 reserved． byJournal∥Software．Allrights 基于聚类分解的高维度量空间索引B+．Tree· 张军旗，周向东+，王梅，施伯乐 (复旦大学计算机与信息技术系，上海200433) Cluster BasedDimensionalMetric IndexB+-Tree SplittingHigh Space ZHANG Bai—Le Mei，SHI Jun-Qi，ZHOUXiang-Dong+，WANG of andInformation University，Shanghai200433，China) (DepartmentComputing Technology，Fudan +Correspondingauthor：E—mail：xdzhou@fudan．edu．cn BL．Clusterbaseddimensionalmetric indexB+-Tree． ZhangJQ，ZhouXD，WangM，Shi splittinghigh space Journal cluster isoftenusedto the Abstract：Inorderto the approach estimatedata improvequeryefficiency,K-means distributioninthecontextof dimensionalmetric index．Butin of space work，theparameters high previous are selected tosomeheuristicmanner．This anew dimensional clusteringusually according paperpresentshigh index based dimensional cluster data is approach--clusterhigh B+．tree．Throughsplining，thespace splitting more toreducethecostofdataaccess．The betweenclusterandthe costis partitioned relationship query finely basedonthe cost formulasto ofthe discussed，and model，thisgive compute query paper