规则评新估计划跟数据质量挖掘系统的探究 刘波.pdfVIP

规则评估方法与数据质量挖掘系统的研究 1 1 2 刘 波 ，潘久辉 ，刘佩珊 (1.暨南大学 计算机科学系，广东 广州，510632； 2.广东拓思软件科学园有限公司，广东 广州，510663） 摘要:为了从数据挖掘结果中获得真正有用的规则，探讨了评估规则的度量，提出了度量关 系的新概念，并应用于发现与评估有用规则集的算法中。经过度量函数之间的相关分析，确 定适合的度量，并据此对规则进行兴趣度评估，指导用户对有用规则的选择。此外，为了解 决数据集成系统中存在“垃圾”数据的问题，提出了数据质量挖掘与控制系统框架，将数据 挖掘规则应用到了数据质量管理中。使用经过评估筛选得到的规则，能够检测数据库中异常、 不完整等数据，解决数据集成过程中部分数据质量问题。通过企业信息集成实例，说明了规 则评估及应用方法的有效性。 关键词：规则评估；度量函数；数据质量挖掘；集成系统 中图分类号：TP302 文献标识码：A Study of rule evaluation method and data quality mining system 1 1 2 LIU Bo , PAN Jiu-hui , LIU Pei-shan (1.Department of Computer Science,Jinan University,Guangzhou, Guangdong,510632,China; 2.Guangdong Tops Software Science Park Co. Ltd.,Guangzhou,Guangdong,510663,China ) Abstract ：In order to get actually useful rules from data mining results, it studied measures for evaluating rules, and defined the new concept of measure relationship, which was applied in the algorithm for discovering and evaluating useful rules. By correlative analysis of measure functions, the suitable measures for rule interestingness evaluation can be determined, so as to guide users to choose useful rules. In addition, for solving the problem of “garbage” data in integration systems, the frame of data quality mining and control system was presented, which applied data mining rules in data quality management. Using the evaluated and selected rules, the abnormal, incomplete data can be detected, so as to solve some data quality problems in data integration. The example of an enterprise information integration shows the effectiveness of the rules’ evaluation and the application method. Key words ：rule evaluation; measure function; data quality mining; integration system 基金项目：国家863 计划项目(20