基于核函数的成分数据缺失值处理

发布时间：2018-10-17 10:17

【摘要】：由于人们的科学意识不断进步,分析研究的科学精神逐渐深入人心,现代生活中常常需要面对数据的收集与处理,以便更高效地完成日常工作。在所有可能出现的数据中,成分数据是一种满足特殊性质的复杂多维数据,一般用于研究一个整体中各部分间关于指定因素下的比例关系。随着经济发展水平不断提高,各行各业越来越意识到精确数据统计带来的好处,成分数据因此也应用得越来越广泛。然而实际问题中,我们发现收集统计的数据常常会存在缺失,例如问卷中的无效或空白信息,收集中的遗漏等等都会产生缺失数据。统计质量会受到缺失数据的影响,导致估计偏差,产生不良结果。故而我们希望数据能够完整,因此对缺失数据的补全显得尤为重要。目前国内外在缺失数据的处理方面已有不少成果,本文在前人的研究基础上,尝试利用核函数的方法进行缺失值填补,研究对比不同方法的优劣。本文分为五章:第一章说明了本文的研究意义,阐述了当前的研究背景,国内外的研究现状,并对一些基本情况作了概述。第二章简要叙述了成分数据的基本概念,以及需要用到相关的相关知识,对研究过程中的大致操作进行描述,并对已有的一些方法给予介绍。第三章是本文重点,提出了基于核函数的几种成分数据缺失值填补法,阐明了提出方法的原因、过程以及具体实现步骤。第四章通过对提出的几种基于核函数的缺失值填补方法与已有常见方法的模拟实验对比,得出实验结果,并对真实数据进行实例分析,以验证方法的可行性。最后一章进行了总结,提炼本文的研究结论,以及对今后研究的展望。
[Abstract]:Due to the continuous progress of people's scientific consciousness, the scientific spirit of analysis and research has gradually taken root in the hearts of the people. In modern life, it is often necessary to face the collection and processing of data in order to complete daily work more efficiently. Among all the possible data, the component data is a kind of complex multidimensional data which satisfies the special properties. It is generally used to study the proportional relationship between the parts of a whole under the specified factors. As the level of economic development continues to improve, various industries are increasingly aware of the benefits of accurate data statistics, and component data are therefore more and more widely used. However, in practical problems, we find that the data collected from statistical data are often missing, such as invalid or blank information in the questionnaire, missing information in the collection and so on. The statistical quality will be affected by the missing data, resulting in the estimation deviation and bad results. Therefore, we want the data to be complete, so it is very important to complete the missing data. At present, there have been a lot of achievements in the processing of missing data at home and abroad. On the basis of previous studies, this paper attempts to use the kernel function method to fill the missing value, and to study and compare the advantages and disadvantages of different methods. This paper is divided into five chapters: the first chapter explains the significance of the research, describes the current research background, domestic and foreign research status, and gives an overview of some basic conditions. The second chapter briefly describes the basic concept of component data and the need to use relevant knowledge to describe the general operation of the research process and to introduce some existing methods. The third chapter is the focus of this paper. Several methods based on kernel function are proposed to fill the missing values of component data. The reason, process and implementation steps of the proposed method are explained. In chapter 4, the experimental results are obtained by comparing the proposed missing value filling methods based on kernel functions with common methods, and the real data are analyzed by an example to verify the feasibility of the method. The last chapter summarizes the conclusion of this paper and prospects for future research.
【学位授予单位】：山西大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：O212.1;F224

【参考文献】