云环境下聚类分解的高维数据混合索引方法  

A new hybrid index method based on cluster splitting under the cloud environment

作　　者：王倩[1] 朱变[1] 

机构地区：[1]周口师范学院计算机科学与技术学院,河南周口466001

出　　处：《周口师范学院学报》2015年第2期116-119,共4页Journal of Zhoukou Normal University

基　　金：河南省科技厅项目(No.2010B520035)

摘　　要：针对云计算环境下分布式存储系统的数据索引不支持复杂查询的问题,笔者提出了云环境下聚类分解的高维数据混合索引方法.首先,采用聚类分解方法对分割数据建立树状索引;然后,以叶节点为单位,通过扫描线算法来获取节点内部所有对象的局部最近邻结果;最后,依据计算的结果得出启发式的裁剪距离.在单节点最近邻计算中,第二个阶段获取外部的最近邻对象采用范围查询算法.实验分析表明,在查询效率上该索引方法高于单纯的聚类方法.与M-tree、顺序查找、iDisance相比,基于聚类分解的混合索引方法在高维查询模式下具有良好的查询效率和负载均衡.Data index distributed storage system in the cloud computing environment does not support complex query problem.The paper presents a new hybrid index method based on cluster splitting.Under the cloud environment,the data space is partitoned more finely to reduce the cost of data access and a tree index.And then the query is performed within each leaf node by using plane-sweeping algorithm,in which all objects,as well as their nearest neighbor candidates,reside in current leaf node,and then an enhanced range query is performed to identify nearest neighbor and discard false alarms by retrieving objects that locate in other leaf nodes,based on the distance between objects and its might-be nearest neighbor of the first step.Experiment results show the method based on cluster splitting is higher than the clustering method alone and the new hybrid index method based on cluster splitting in high dimensional query mode is better than iDistance,M-Tree and sequence scan.

关 键 词：云计算 聚类分解 混合索引 高维查询 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]