【作者】 张松富；

【导师】 李卫华；

【作者基本信息】 广东工业大学 ， 计算机技术， 2007， 硕士

【摘要】 如何提高数据维护能力，保证数据质量是摆在数据库应用方面的重要课题。对于数据质量问题，数据库用户大多是从企业行业内部管理制度角度出发，加强制度建设，增加人员责任心等措施来提高数据质量。数据库厂商则从数据库技术方面力求提供通用产品解决数据质量，但其提供的方法大多是基于技术层面的，缺乏与具体用户复杂业务逻辑的有效结合。国内一些研究机构也提出了关于解决数据质量的理论，他们认为数据质量管理如同产品质量管理一样贯穿于数据生命周期的各个阶段，但目前尚缺乏一个系统的思路。国内外大多侧重在数据仓库和数据挖掘等系统范畴下的于数据质量问题，属事后型。研究海量数据进行清洗比较多，但这些方法不能直接提高数据本身的质量，只能减少不可靠数据的对决策的影响，不能从根本上解决问题。对于数据质量的监控，对数据维护的投入应该前移到数据的采集、繁衍期。数据质量通用标准是：数据的完整性、一致性、准确性。对于单源数据信息系统要提高数据的质量有一定难度，尤其是数据的准确性。本文提出对于无法从逻辑上判断数据准确性时，通过引入数理统计理论和方法来判断数据准确的准确程度，从而发现问题数据。数据维护通常是在数据出现问题后，通过信息管理系统或数据库管理系统对数据进行调整、修改以维护其正常状态。这种模式使数据维护始终处于被动状态，本文引入Agent技术，把数据维护工作融入系统正常运行，实时进行问题数据的查找，主动发现问题并用规范的方法来处理问题数据，使数据维护工作效率得到提高，信息系统平均无故障时间增长。本文结合大集中模式下税收业务管理系统，阐述了智能数据维护的实现过程。在策略方面，笔者提出主动的数据维护模式，改变数据维护思路。将数据维护工作贯穿到系统日常运行过程中，使数据维护量不会随时间增加而呈线性增长的趋势，使系统处在一种边运行、边“润滑”的状态，可以增长系统的平均无故障时间，较大地提高系统效率。问题数据的查找提前，且处理方法是标准可行的，使数据维护工作量处于可预测、可控的状态。数据维护纳入系统运行的范畴，形成闭环反馈系统，改变数据维护的概念。对于数据维护系统运行结果的处理，则是对现有系统的一个客观、量化的评价。可以在不增加业务逻辑的复杂程度的前提下，改善系统运行的效果。这对现存系统是一个补充。本文所采用的策略和方法应该对信息系统的数据维护有着可以借鉴的通用性。更多还原

【Abstract】 How to improve the data maintenance, data quality assurance are important issues placed in a database application. For data quality issues, database users mostly strengthen the construction of systems, increase staff accountability, and take other measures to improve data quality from the enterprise internal management systems perspective. Database technology manufacturers strive to provide common data quality product solutions, provide most of the methods based on the technical level, which lacks of specific users with complex business logic of the effective combination. Some domestic research institutes have also raised the quality of data on the settlement of the theory. They believe the data quality management as the same product quality management throughout the life cycle of data in all phases. However, at present it lacks of a system of thinking. Most domestic and international researchers focus on the data warehouse and data mining systems in the areas of data quality issues, which were after the event techniques. There are many massive data cleaning researches, but they cannot directly improve the quality of the data itself. They can only reduce unreliable data to affect the decision-making and fail to resolve the fundamental problem. For data quality monitoring and maintenance of data, effort should be moved to before the acquisition, reproduction period.Common standards of data quality are: data integrity, consistency and accuracy. For single-source data information systems to improve the quality of the data is difficult, particularly in the accuracy of the data. When we cannot judge the accuracy of the data from the logic, this thesis introduces mathematical statistical methods to determine the accuracy of data accuracy, thereby discovering data problems.Data Maintenance is usually after the data occurs problems, through information management systems or database management system for data adjustment and revisions to preserve its normal state. This format enables the data maintenance always in a passive position. This thesis introduces Agent technology to set data maintenance into the system normal operation. It finds data problems in real-time, take the initiative to find the problem and using standardized methods to deal with the data problems. This increases data maintenance work efficiency and MTBF (Mean time Before Failure) of Information System.This thesis, based on CTAIS (China Tax Administrator Information System), describes the implementation process of intelligent data elaborated safeguard. At a strategic level, the author offers active data maintenance mode, change data maintenance ideas. Data Maintenance will work through the day-to-day operation of the system process. Prevent maintaining data work from increasing over time in linear growth trends. Allow the system "lubricant" while running. It will increase the average time to failure, and greatly improve efficiency. Identify data problems early and the approach both standard and feasible. Data maintenance workload is at a predictable, controllable condition. Data maintenance sets in system operations and forms a closed-loop feedback system, changing data maintenance concept. For the results of the operation of data maintenance system, it is an objective, quantitative evaluation on the existing system. Under the premise of not increase the complexity of business logic level improvs the operating results. It is a good complement of the existing system. The strategy used and the method of information systems for data maintenance has a useful generalization.更多还原