降维空间视觉认知增强的多维时变数据可视分析方法  

Visual Analytics for Multidimensional Time-Varying Data via Dimension Reduced Visual Perception

作　　者：周志光 汤成[3] 刘玉华 刘亚楠 许建龙 Zhou Zhiguang;Tong Cheng;Liu Yuhua;Liu Yanan;Xu Jianlong(School of Information and Engineering,Zhejiang University of Uinance ＆ Economics,Hangzhou 310018;State Key Laboratory of CAD＆CG Zhejiang University Hangzhou 310058;School of Information,Zhejiang Sci-Tech University Hangzhou 310018)

机构地区：[1]浙江财经大学信息管理与工程学院,杭州310018 [2]浙江大学CAD&CG国家重点实验室,杭州310058 [3]浙江理工大学信息学院,杭州310018

出　　处：《计算机辅助设计与图形学学报》2018年第7期1194-1204,共11页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61303133);浙江省自然科学基金(LY18F020024);全国统计科学研究重大项目(2015LD03);浙江大学CAD&CG国家重点实验室开放课题(A1806);浙江省一流学科A类(浙江财经大学统计学)资助

摘　　要：针对多维时序数据可视分析过程中降维算法表现出的局限性,提出一种降维空间视觉认知增强的多维时序数据可视分析方法.在多维标度算法的基础上,通过正交变换使不同时间步投影点的偏移最小化,帮助用户对感兴趣的时间模式进行有效的视觉认知及追踪;为避免投影点之间的相互遮挡,引入六边形对投影空间进行划分,增强用户对降维空间特征的视觉认知和交互;进而引入层次聚类方法对投影点进行聚类分析,帮助用户快速感知多维数据之间的关联关系;最后设计面向聚类特征时序演变的分组动画策略,突出相邻时间聚类特征的演化特点和时序模式.集成上述可视化方法,开发面向多维时序数据可视分析原型系统,通过经济统计数据、空气质量监测数据的实例分析,进一步验证该系统的有效性和实用性.In this paper, we propose a visualization system for multidimensional time-varying data analysis via dimension reduced visual perception, which enables users to better perceive the temporal features of the time-varying multidimensional data. Firstly, the orthogonal transformations are conducted to minimize the offsets of MDS plots between adjacent time steps, which do great favors for the visually tracking of temporal patterns of interest and maintaining an accuracy mental map. Then, hexagons are applied to tessellate the plane to eliminate the overlap in the resulting points and facilitate the visual perception and interaction of MDS plots. Furthermore, we employ a hierarchical clustering algorithm and design a specialized glyph to enhance the visual perception of temporal clusters. Finally, a cluster based grouping animation is designed to highlight the temporal patterns between adjacent timestamps. By integrating the above visualization methods, a time-varying multidimensional data visualization analysis prototype system is developed. We demonstrate the effectiveness and usefulness of TMDS in case studies with the economic data and air quality data.

关 键 词：多维标度 时序分析 数据可视化 层级聚类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]