【摘要】：雙標(biāo)圖是一種廣泛應(yīng)用的可視化分析方法,但是當(dāng)所研究的數(shù)據(jù)包含較多變量時,如果直接用雙標(biāo)圖進(jìn)行分析會導(dǎo)致圖中較多變量重疊,不能很清晰地觀察變量間的相關(guān)關(guān)系,可視化程度較低,分析效果不精確,因此尋找一些能夠有效解決一般的多變量數(shù)據(jù)的統(tǒng)計方法就非常必要.針對上述問題,本文提出了兩種增強雙標(biāo)圖的可視化的分析方法,第一種是基于聚類分析的雙標(biāo)圖分析方法,首先通過對原始數(shù)據(jù)進(jìn)行聚類分析,得到新的數(shù)據(jù)集,然后對得到的新數(shù)據(jù)集進(jìn)行雙標(biāo)圖分析.另一種方法是基于主成分和聚類分析提出一種新的雙標(biāo)圖分析方法.此兩種方法不僅保留了數(shù)據(jù)間的絕大多數(shù)信息,而且使得雙標(biāo)圖的可視化程度增強.對兩種新的雙標(biāo)圖方法進(jìn)行實證分析,并與原始數(shù)據(jù)構(gòu)成的雙標(biāo)圖進(jìn)行比較研究,驗證了該方法的有效性,最后將兩種新的雙標(biāo)圖方法推廣應(yīng)用到成分?jǐn)?shù)據(jù)上.論文主要由五章組成.第一章是引言,主要介紹了本文的研究背景,問題的提出及其實際意義,簡要說明本文的工作及創(chuàng)新之處,并給出了本文的主要結(jié)構(gòu).第二章是雙標(biāo)圖的簡介,對雙標(biāo)圖的一般模型進(jìn)行了描述,簡單介紹了雙標(biāo)圖的基礎(chǔ)理論知識,并簡單介紹了三種類型的雙標(biāo)圖.第三章簡紹了兩種增強雙標(biāo)圖可視化的方法.針對多變量數(shù)據(jù)集,如果直接用雙標(biāo)圖進(jìn)行分析會導(dǎo)致圖中較多變量重疊,不能很清晰地觀察變量間的相關(guān)關(guān)系,可視化程度較低,分析效果不精確,故本章提出了兩種增強雙標(biāo)圖的可視化的分析方法.第一種是基于聚類分析的雙標(biāo)圖分析方法,首先對原始數(shù)據(jù)集進(jìn)行分類,得到一些新的數(shù)據(jù)集,然后利用雙標(biāo)圖對新的數(shù)據(jù)集進(jìn)行分析,分析每類中原始變量與均值變量之間的關(guān)系.對新的雙標(biāo)圖分析方法進(jìn)行實例分析,并與原始數(shù)據(jù)構(gòu)成的雙標(biāo)圖進(jìn)行比較研究,驗證了該方法的有效性.第二種是基于聚類分析和主成分分析的雙標(biāo)圖分析方法,首先基于主成分分析和聚類分析,對原始數(shù)據(jù)集進(jìn)行分類,得到新的數(shù)據(jù)集,對新的數(shù)據(jù)集進(jìn)行雙標(biāo)圖方法進(jìn)行了實例驗證,驗證了該方法的有效性.以上兩種方法不僅保留了數(shù)據(jù)間的絕大多數(shù)信息,而且使得雙標(biāo)圖的可視化程度增強.第四章介紹了成分?jǐn)?shù)據(jù)雙標(biāo)圖的構(gòu)造步驟及其成分?jǐn)?shù)據(jù)的基本理論,將第三章提出的兩種方法應(yīng)用到成分?jǐn)?shù)據(jù)中進(jìn)行實例驗證.第五章是結(jié)論部分.本文對兩種增強雙標(biāo)圖可視化分析方法進(jìn)行了總結(jié),發(fā)現(xiàn)在多變量數(shù)據(jù)集條件下,直接利用傳統(tǒng)的雙標(biāo)圖分析方法存在一些弊端,即可視化可能會降低,而本文提出的這兩種增強雙標(biāo)圖可視化的分析方法很好的解決了雙標(biāo)圖可視化低的問題.本文的目的是希望找到一種既不丟失數(shù)據(jù),又能很好的分析多變量數(shù)據(jù)集的雙標(biāo)圖分析方法,使得可視化增強.
[Abstract]:Double plot is a widely used visual analysis method, but when the data under study contains more variables, if the data is analyzed directly, it will lead to more variables overlap in the graph, so the correlation between variables can not be observed clearly. The degree of visualization is low and the analysis effect is not accurate. Therefore, it is necessary to find some statistical methods that can effectively solve the general multivariable data. In order to solve the above problems, this paper proposes two methods to enhance the visualization of double plot. The first method is based on clustering analysis. Firstly, a new data set is obtained by clustering the original data. Then the new data set is analyzed by double plot. Another method is a new method based on principal component and cluster analysis. These two methods not only retain most of the information between the data, but also enhance the visualization degree of the double plot. Two new double mapping methods are empirically analyzed and compared with those of original data. The validity of this method is verified. Finally, two new double mapping methods are extended to component data. The thesis consists of five chapters. The first chapter is the introduction, which mainly introduces the research background, the problem and its practical significance, briefly explains the work and innovation of this paper, and gives the main structure of this paper. The second chapter is a brief introduction of double plotting. The general model of double plotting is described, the basic theoretical knowledge of double plotting is briefly introduced, and three types of double plotting are briefly introduced. In the third chapter, two methods to enhance the visualization of double map are introduced briefly. In view of multivariate data sets, if the analysis of multivariate data sets is carried out directly, it will lead to the overlapping of more variables in the graph, so the correlation between variables can not be observed clearly, the visualization degree is low, and the analysis effect is not accurate. Therefore, this chapter proposes two methods to enhance the visualization of double maps. The first method is based on clustering analysis. First, the original data set is classified and some new data sets are obtained, then the new data set is analyzed by using double scale graph. The relationship between the original variable and the mean variable in each class is analyzed. A case study of the new double plot analysis method is carried out and compared with that of the original data. The validity of the method is verified. The second method is based on cluster analysis and principal component analysis. Firstly, based on principal component analysis and clustering analysis, the original data sets are classified and a new data set is obtained. An example is given to verify the validity of the new method. The above two methods not only retain most of the information between the data, but also enhance the visualization of the two maps. In chapter 4, the construction steps and the basic theory of the composition data are introduced. The two methods proposed in chapter 3 are applied to the component data for example verification. Chapter five is the conclusion. In this paper, two methods of enhanced double map visualization analysis are summarized. It is found that under the condition of multivariable data set, there are some disadvantages in using traditional double map analysis method directly, that is, visualization may be reduced. The two analysis methods proposed in this paper can solve the problem of low visualization of double diagrams. The purpose of this paper is to find a bivariate map analysis method that can analyze multivariate data sets without losing data, so as to enhance visualization.
【學(xué)位授予單位】：山西大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：O212

【參考文獻(xiàn)】

相關(guān)期刊論文 前7條

1 陳享光;孫科;;我國行業(yè)間工資差距的動態(tài)考察[J];中國人民大學(xué)學(xué)報;2014年02期

2 嚴(yán)威凱;;雙標(biāo)圖分析在農(nóng)作物品種多點試驗中的應(yīng)用[J];作物學(xué)報;2010年11期

3 薛付忠;王潔貞;郭亦壽;胡平;;人類群體遺傳結(jié)構(gòu)的雙標(biāo)圖模型及其應(yīng)用[J];科技導(dǎo)報;2006年05期

4 嚴(yán)威凱,盛慶來,胡躍高,L A Hunt;GGE疊圖法─分析品種×環(huán)境互作模式的理想方法[J];作物學(xué)報;2001年01期

5 孟憲偉,杜德文,吳金龍;成分?jǐn)?shù)據(jù)的因子分析及其在地質(zhì)樣品分類中的應(yīng)用[J];長春科技大學(xué)學(xué)報;2000年04期

6 周蒂;地質(zhì)成分?jǐn)?shù)據(jù)統(tǒng)計分析——困難和探索[J];地球科學(xué);1998年02期

7 張崇甫,陳述云;成分?jǐn)?shù)據(jù)主成分分析及其應(yīng)用[J];數(shù)理統(tǒng)計與管理;1996年04期

，

本文編號：2413634