CSSAQP:一种基于聚类的分层抽样近似查询处理算法  

CSSAQP:An Approximate Query Algorithm Based On Clustering Stratified Samping

作　　者：谢金星[1] 李晖[1] 陈梅[1] 戴振宇[1] XIE Jinxing;LI Hui;CHEN Mei;DAI Zhenyu(Guizhou Engineering Lab for ACMIS,Guizhou University,Guiyang 550025)

机构地区：[1]贵州大学先进计算与医疗信息服务工程实验室,贵阳550025

出　　处：《计算机与数字工程》2017年第6期1121-1126,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61462012,61562010,U1531246);基于云计算的医疗信息管理系统关键技术研究及应用(编号:GY[2014]3018);贵州省重大应用基础研究项目(编号:JZ20142001);贵州省教育厅自然科学项目(编号:黔科合人才团队字[2015]53号);贵州大学研究生创新基金(院级)资助

摘　　要：近似查询处理技术常被应用于海量数据的多维分析,以缩短查询执行的时间,同时返回尽可能准确的结果。由于海量数据中常存在许多极端值,会严重影响近似查询处理的结果。因此针对海量数据的聚集操作,论文提出CSSAQP算法,先将原始数据集按某一数值列直观的聚为三类,分别代表大值簇、小值簇和常值簇,再对各簇按分组属性分别进行分层抽样,构建总体样本集,最后通过查询重写在总体样本集上执行查询,以缩短海量数据聚集操作的查询时间,同时提高查询任务的准确性。通过实验验证,证明了该算法不仅可以缩短聚集查询的时间,同时还能有效提高查询结果的精度。The approximate query processing technique is often applied to multidimensional analysis of massive data to short?en the execution time of the query and return the results as accurate as possible.Because of many extreme values in massive data,itwill seriously affect the results of approximate query processing.Therefore,for the aggregation of massive data,this paper proposesa algorithm CSSAQP,which first clustered the original data set into three categories by a column,representing large clusters,smallclusters and constant clusters,then use stratified sampling for each cluster by the group attribute,and constructed the overall sam?ple,finally,the query is rewritten on the overall sample set to reduce the query time of the massive data aggregation operation,andimprove the accuracy of the query task.Experiments show that the algorithm can not only shorten the time of aggregation query,butalso improve the accuracy of query results.

关 键 词：近似查询处理 聚集查询 聚类 分层抽样 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]