联机分析挖掘技术及其应用研究-智能与分布计算实验室

联机分析挖掘技术及其应用研究

姓名	王小宜
论文答辩日期	2004.05.10
论文提交日期	2004.05.11
论文级别	硕士
中文题名	联机分析挖掘技术及其应用研究
英文题名	The Research of On Line Analytical Mining Technology and its Application
导师1	卢正鼎
导师2
中文关键词	联机分析处理;数据挖掘;联机分析挖掘;事务优化;离群数据挖掘
英文关键词	On Line Analytical Processing;Data Mining;On Line Analytical Mining;transaction optimization;outlier data mining
中文文摘	联机分析挖掘技术融合了联机分析处理技术和数据挖掘技术，成为决策支持应用系统新的技术依托。该技术的研究目前处于起步阶段，关于联机分析挖掘技术的基本原理、关键技术、系统模型以及应用开发技术等问题还没有系统的研究。合理、高效的系统体系结构是联机分析处理技术与数据挖掘技术完美融合的保证。联机分析挖掘系统的体系结构在于两个层面：概念和逻辑。基于这两个层面，给出了联机分析挖掘系统的概念模型和逻辑模型，为联机分析挖掘系统的开发提供了理论的指导。结合实际项目背景给出国家外汇信息管理决策原型系统的系统架构。联机分析挖掘系统是由事务驱动的，事务的优化能提高系统性能。根据联机分析挖掘事务模型，事务间的协作成为联机分析挖掘事务优化的主要技术之一。运用导出对象作为事务协作的“应标”对象，改进了“招应标”事务协作模型。改进后的模型采用代理池的协作管理模式，并将数据传输与协作匹配相分离，提高了协作事务间的数据传输效率。联机分析挖掘技术对数据挖掘算法提出了新的要求。根据新的需求，联机分析挖掘系统中的数据挖掘应该充分利用多维立方体以提高挖掘速度和挖掘能力，并给出联机分析挖掘系统中数据挖掘的三种策略。离群数据挖掘是数据挖掘研究的一个重要分支。根据课题背景，给出一个针对时序数据的离群数据挖掘算法的改进算法。该算法基于小波理论对数据进行简化，减少了计算复杂度。在上述研究与分析的基础上，作为主要成员设计实现了基于联机分析挖掘技术的国家外汇信息管理决策原型系统中的数据挖掘部件。该部件集成了多种数据挖掘方法，能有效的对数据进行分析挖掘。
英文文摘	On Line Analytical Mining integrates On Line Analytical Processing and Data Mining technology. These years it becomes an important new technology applied in Decision Support System. But the research on the basic theory, key technology, system model and technology of this aspect are not systematic. Reasonable and efficient system architecture ensures the perfect integration of On Line Analytical Processing and Data Mining. On Line Analytical Mining system architecture lies in concept model and logic model. These two models provide a theoretic guide in the implement of On Line Analytical Mining system. Based on the project background, the system architecture of SAFE-MIDSS is put forward. On Line Analytical Mining system is driven by transaction, so the optimization of transaction can improve system performance. According to On Line Analytical Mining transaction model, the cooperation of sub-transactions becomes one of the most important technologies to optimize the transaction. Using export data object, an improved the model to cooperate sub-transactions is put forwards. The improved model brings forward a thinking of cooperation management based on agent pool and separates data transaction from cooperation matching. As a result, it improves the data transport efficiency. The applications of On Line Analytical Mining bring a new request on the data mining algorithm. To improve mining ability, data mining should utilize the multi-dimension data. Three policies are given out to use the data mining algorithm in On Line Analytical Mining system. Outlier data mining is an important embranchment in data mining research. Based on the project background, an improved outlier data mining algorithm for time series data is given out. Based on the wavelet, the time to process data is shortened. A decision support system called SAFE-MIDSS is designed and implemented, based on technology and methods above. A detailed implementation of its data mining component is specified. It integrates several data mining method and realizes to analyze and find knowledge in data.