多元网络可视分析综述

A Survey on the Visual Analytics of Multivariate Networks

多元网络是一种常见的网络结构,不仅描述了数据之间复杂的关系结构,而且记录了数据本身所具有的多维属性信息.多元网络可视分析方法研究能够帮助用户全面而深入地探索多元网络中隐藏的复杂特征及其关联模式.文中面向多元网络可视分析方法进行深入的调研和分析,首先从可视化展现形式出发,介绍多视图协同、简化表达、属性布局、点边映射和矩阵表达等多元网络可视化方法;其次对多元网络可视分析中的交互方式进行层级划分和详细介绍,包括视图层级、视觉结构层级和数据层级;进一步概述多元网络可视分析方法在社交网络、地理交通和深度学习等领域的应用;最后总结多元网络可视分析研究面临的挑战,对其未来发展趋势进行展望.

As a common data structure,multivariate network does not only describe complex topological relationships among nodes with edges,but also record various attributes associated with nodes.A large number of visual analytics methods have been proposed enabling users to get deeper insights into the complex features and association patterns of multivariate networks.In this paper,we give a survey on the visual analytics of multivariate networks.Firstly,we introduce a variety of presentations for multivariate networks,such as coordinated views,graph abstraction,attribute-driven layout,on-node/edge encoding and graph matrix.Then,a set of interactions designed for the exploration of multivariate networks are classified into three categories,including the view level,visual-structure level and data level.Furthermore,we investigate the applications of visual analytics of multivariate networks in a variety of domains including social network,deep learning,etc.Finally,the challenges and future work of multivariate network visualizations are discussed and concluded.