Abstract: To repair error or inconsistent data produced in the operations of enterprise's multiple sources databases effectively and automatically,a new repairing algorithm based on functional dependencies and inclusion dependencies was presented.For violations on functional dependencies,the related attribute statistical measures were computed by this algorithm and the tuples to be modified by tuples'confidence were selected.Aiming at the violations on inclusion dependencies,parts of attribute values between different datasets were matched and the method to update or insert new tuples was determined.Deletion operations were not adopted in the algorithms for no loss of information in original databases.The proposed algorithm had features such as objective,accurate and efficient,and could solve the inconsistent problems in enterprise information integration.

Key words: information integration, consistency, repair, database

摘要： 为了有效、自动地修复在企业多个信息源的数据库操作中产生的错误或不一致数据,基于数据库的函数依赖和包含依赖,提出新的修复算法。算法针对违背函数依赖的数据,计算相关属性的统计度量,根据元组的可信度选择需要修改的元组;针对违背包含依赖的数据,匹配不同数据集之间的部分属性值,确定如何修改或插入新元组。算法对数据库不实施删除操作,保证了原数据库信息的完整性,具有客观、准确、高效等特性,能够应用于解决企业信息集成中出现的数据不一致问题。

关键词: 信息集成, 一致性, 修复, 数据库