Aggregate Query Processing on Incomplete Data Based on Denotational Semantics

DOI：10.13328/j.cnki.jos.005876

 作者 单位 E-mail 张安珍 哈尔滨工业大学 计算机科学与技术学院, 黑龙江 哈尔滨 150001沈阳航空航天大学 计算机学院, 辽宁 沈阳 110000 azzhang@hit.edu.cn 李建中 哈尔滨工业大学 计算机科学与技术学院, 黑龙江 哈尔滨 150001 高宏 哈尔滨工业大学 计算机科学与技术学院, 黑龙江 哈尔滨 150001

本文研究了基于符号语义的不完整数据聚集查询处理问题.不完整数据又称为缺失数据，缺失值包括可填充的和不可填充的两种类型.现有的缺失值填充算法不能保证填充后查询结果的准确度，为此，本文给出不完整数据聚集查询结果的区间估计.本文在符号语义中扩展传统关系数据库模型，提出一种通用不完整数据库模型，该模型可以处理可填充的和不可填充的两种类型缺失值.在该模型下，提出一种新的不完整数据聚集查询结果语义:可靠结果.可靠结果是真实查询结果的区间估计，可以保证真实查询结果很大概率在该估计区间范围内.本文给出线性时间求解SUM、COUNT和AVG查询可靠结果的方法.真实数据集和合成数据集上的扩展实验验证了本文所提方法的有效性.

This paper studies the problem of aggregate query processing over incomplete data based on denotational semantics. Incomplete data is also known as missing values and can be classified into two categories:applicable nulls and inapplicable nulls. Existing imputation algorithms cannot guarantee the accuracy of the query result after imputation. We give the interval estimation of the aggregate query result. This paper extends the relational model under the denotational semantic, which can cover all types of incomplete data. We define a new semantic of aggregate query answers over incomplete data, reliable answers. Reliable answers are interval estimations of the ground-truth query results, which can cover the ground-truth results with high probability. For SUM, COUNT and AVG queries, we propose linear approximate evaluation algorithms to compute reliable answers. The extended experiments on the real datasets and synthetic datasets verify the effectiveness of the method proposed in this paper.
HTML  下载PDF全文  查看/发表评论  下载PDF阅读器