【作者】 刘强；

【导师】 聂斌；

【作者基本信息】 天津大学 ， 工业工程， 2012， 硕士

【摘要】 经过几十年的发展,统计过程控制研究已经成为一个综合性的质量管理理论体系,数据处理技术也得到了很大提高,但是在一些技术领域还不完善,在检测精度、方法创新以及其他技术层面还有待提高。本人提出一种基于聚类分析的变点检测方法,该方法在聚类分析的基础上对变点检测的方法进行了改进,在检测精度上面有了一定的提高。本文首先是对聚类分析技术和判定方法进行了阐述,聚类分析的目的是将原始数据归类,方便变点的检测;判定规则是将变点检测的技术指标等因素加以阐述,为变点的检测设定必要的规则。其次,本文将以上技术进行了实际操作,该技术的成功实施证明了该技术的可行性。最后,在第五章本文将提出的技术和其他两项技术进行了不同程度的比较,利用犯两类错误的概率将本文提出的技术和其他两类技术进行了比较,进一步验证了该理论方法的优越性能。本文的创新点是聚类分析和数据密度函数的创新应用,这两项技术的结合应用能够在一定程度上将变点检测的精度保持在一个比较高的水平,这也是本文提出的技术方法的意义所在。在文章末尾,本文将该项技术进行了总结分析,将该技术的优缺点进行了具体陈述,得出了一些值得借鉴的结论,并对该项技术未来的研究方向做出了一些建议和展望。更多还原

【Abstract】 After developed about several decades, statistical process control has became a comprehensive quality control theory, with advanced technique in data handling, but it has to be improved in some technique fields like detect accuracy、new methods and other techniques。I suggest a change point detection methods based on clustering analysis,it could improve the change point detection method by clustering analysis, that is good for improving detect accuracy. This article first state the technique of clustering analysis and detection rules to classify the initial data for change point detection. The detection rules technique in this article describes the characters used in change point detection, it could set the rules for change point detection. In the second, we formulate the characters of the method in change point detection to approve the feasibility of the above technique. In the last, we compare the above technique with two other methods in several fields to approve the advanced function of the technique in this article. The innovation of this article is the comprehensive technique of clustering analysis and detection rules, this comprehensive technique improve the change point detection accuracy, and this is also importance of this article. In the last of this article, we made a conclusion about this technique, it include the merits and feedbacks of this technique, so this conclusion is worth of reference,we also give the future development and some suggestions of this technique.更多还原