基于可能世界模型的关系数据不一致性的修复

发布时间：2018-06-19 12:50

本文选题：不一致性 + 函数依赖　；参考：《软件学报》2016年07期

【摘要】：针对关系数据的不一致性虽然已有各种修复方法被提出,但这些修复策略在构建最终修复方案的过程中只分析函数依赖包含属性的信息(即,数据集的部分信息),且偏向于修复代价最小的方案,而忽略了数据集的其他属性以及这些属性与函数依赖包含属性之间的相关性.为此,提出一种基于可能世界模型的不一致性修复方法.它首先构造可能的修复方案,然后从修复代价和属性值相关性两个方面量化各个候选修复方案的可信性程度,并最后找出最优的修复方案.实验结果验证了所提出的修复方法取得了比现有基于代价的修复方法更好的修复效果.同时也分析了错误率和不同类型概率量化对所提出的修复方法的影响.
[Abstract]:Although a variety of repair methods have been proposed for the inconsistency of relational data, these repair strategies only analyze the information of functional dependencies containing attributes in the process of building the final fix (that is, The partial information of the data set is partial, and it tends to repair the scheme with the least cost, while ignoring the other attributes of the data set and the correlation between these attributes and the functional dependency containing attributes. For this reason, a method of inconsistency repair based on the possible world model is proposed. It first constructs a possible repair scheme, then quantifies the credibility of each candidate repair scheme from two aspects of repair cost and attribute value correlation, and finally finds out the best repair scheme. The experimental results show that the proposed method is more effective than the existing cost-based methods. At the same time, the effects of error rate and different types of probability quantization on the proposed restoration methods are also analyzed.
【作者单位】：西北工业大学计算机学院;
【基金】：国家重点基础研究发展计划(973)(2012CB316203) 国家自然科学基金(61332006,61472321,61502390) 西北工业大学基础研究基金(3102014JSJ0013,3102014JSJ0005)~~
【分类号】：TP311.13

【参考文献】