IEEE Trans Vis Comput Graph. 2017 Jan;23(1):251-260. doi: 10.1109/TVCG.2016.2598497.
Data flow systems allow the user to design a flow diagram that specifies the relations between system components which process, filter or visually present the data. Visualization systems may benefit from user-defined data flows as an analysis typically consists of rendering multiple plots on demand and performing different types of interactive queries across coordinated views. In this paper, we propose VisFlow, a web-based visualization framework for tabular data that employs a specific type of data flow model called the subset flow model. VisFlow focuses on interactive queries within the data flow, overcoming the limitation of interactivity from past computational data flow systems. In particular, VisFlow applies embedded visualizations and supports interactive selections, brushing and linking within a visualization-oriented data flow. The model requires all data transmitted by the flow to be a data item subset (i.e. groups of table rows) of some original input table, so that rendering properties can be assigned to the subset unambiguously for tracking and comparison. VisFlow features the analysis flexibility of a flow diagram, and at the same time reduces the diagram complexity and improves usability. We demonstrate the capability of VisFlow on two case studies with domain experts on real-world datasets showing that VisFlow is capable of accomplishing a considerable set of visualization and analysis tasks. The VisFlow system is available as open source on GitHub.
数据流系统允许用户设计流程图，指定系统组件之间的关系，这些组件处理、过滤或可视化呈现数据。可视化系统可能会受益于用户定义的数据流程，因为分析通常包括根据需要呈现多个图形，并在协调的视图中执行不同类型的交互式查询。在本文中，我们提出了 VisFlow，这是一个基于网络的表格数据可视化框架，它采用了一种称为子集流模型的数据流程模型。VisFlow 专注于数据流中的交互式查询，克服了过去计算数据流系统的交互性限制。特别是，VisFlow 应用嵌入式可视化，并支持在面向可视化的数据流程中进行交互式选择、刷选和链接。该模型要求通过流传输的所有数据都是原始输入表的某个数据项子集（即表行的组），以便可以为子集明确分配渲染属性，以进行跟踪和比较。VisFlow 具有流程图的分析灵活性，同时降低了图表的复杂性并提高了可用性。我们通过两个案例研究展示了 VisFlow 的功能，这些案例研究涉及具有实际数据集领域专家的案例，表明 VisFlow 能够完成相当多的可视化和分析任务。VisFlow 系统可在 GitHub 上作为开源使用。
