非对称关系中的数据可视化研究  

Research on data visualization in asymmetric relations

作　　者：陈善为[1] 

机构地区：[1]宝鸡文理学院教育学院,陕西宝鸡721016

出　　处：《信息技术》2016年第6期59-62,共4页Information Technology

基　　金：2014陕西省教育厅资助项目(14JK1021)

摘　　要：基于多维尺度分析(MDS)和自组织映射图(SOM)的二维可视化方法一般都是利用对称距离矩阵对对象在数据集中的关系进行表示和可视化的。然而在如贸易之类的许多实际应用中,对称距离矩阵难以得到,而基于原始数据的非对称相似矩阵却比较容易获取。但是,一般传统方法不能很好的表示非对称矩阵,为了解决这个问题,文中提出一种动态学习算法,用以生成非对称相似数据矩阵,希望通过这种方式,把复杂数据直观化。该方法分别生成两种映射图,每张图都分别拥有和矩阵对应的行向量和列向量。为了对算法有更好的理解,文中采用两种分析工具:聚类分析和距离分析,用以对不同映射图的不同状态进行分析和比较。实践表明,该动态学习方法能极大促进对非对称关系中抽象数据的理解。The 2D visualization method based on multidimensional stealing analysis and self organization map is generally based on the symmetry distance matrix to represent and visualize the relationship between the object and the data set. However, in many practical applications such as trade, symmetric; distance matrix is hard to get, the asymmetric; similarity matrix based on the raw data is easy to obtain. However, the general traditional method can not express the asymmetric matrix well. In order to solve this problem, this paper presents a dynamic; learning method, which is used to generate asymmetric; similarity data matrix, it is hoped that by this way, complex data is visualized intuitively. The method generates two mapping graphs, respectively, and each graph has row vector and column vector corresponding to the matrix. In order to understand this method better, two kinds of analysis tools are used： cluster analysis and distance analysis, to handle different states of different mapping diagrams. The practice shows that this dynamics learning method can greatly promote the understanding of abstract data in asymmetric relations.

关 键 词：数据可视化 非对称关系 聚类分析 距离分析 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]