I-Corps: Filtergraph - A fast and intuitive data visualization interface for massive datasets

I-Corps：Filtergraph - 适用于海量数据集的快速直观的数据可视化界面

• 批准号: 1443314
• 负责人: Keivan Stassun
• 金额: $ 5万
• 依托单位: Vanderbilt University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-07-01 至 2015-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1443314&HistoricalAwards=false
• 关键词: Corps; Filtergraph; fast; intuitive; data

Increasingly in basic research and many other sectors, there is a need for quick-look visualization of large datasets in order to (a) conceptualize the nature of the data, and (b) begin identifying data patterns to guide more computationally costly deep-dive analysis. High-end tools for data basing and graphing exist, but usually entail a large overhead in time and effort associated with: importing the data; correctly specifying meta-data; keeping track of a large number of variable names; issuing and scripting commands for graphing; plotting pairwise variables over many iterations; filtering out bad data; re-rendering plots to restricted data ranges; representing multiple variables at once; and so on. This team has developed Filtergraph, a web based tool that allows rapid and intuitive visualization of massive, multi-dimensional datasets. The web-based tools developed by this team will allow researchers to instantly visualize any dataset in many dimensions at once, and will permit researchers to move through the data rapidly, and quickly filter and slice them. This visual interaction will occur without the need for cumbersome meta-data specification, without the need to remember details of the data structure, and without the need to write scripts to render sophisticated visualizations. The tools developed will be a means to an end: A new, highly visual way of transcending preconceptions about the data and instead let the data speak for themselves. This mechanism would free researchers to creatively and collaboratively engage their expert intuition as they visually and rapidly explore massive datasets for unexpected patterns or outliers that signal important discovery.

在基础研究和许多其他领域中，越来越多地需要快速对大数据集的可视化，以（a）概念化数据的性质，以及（b）开始识别数据模式以指导更多的计算上昂贵的深度潜水分析。 存在用于数据基础和图形的高端工具，但通常需要在与：导入数据相关的时间和精力上进行大型开销；正确指定元数据；跟踪大量可变名称；发行绘画命令和脚本命令；在许多迭代上绘制成对变量；滤除不良数据；重新渲染限制数据范围的图；一次代表多个变量；等等。该团队开发了FilterGraph，这是一种基于Web的工具，可以快速，直观的多维数据集可视化。该团队开发的基于Web的工具将使研究人员立即在许多维度上立即可视化任何数据集，并允许研究人员快速浏览数据，并快速过滤和切片。这种视觉互动将不需要繁琐的元数据规范，而无需记住数据结构的详细信息，也无需编写脚本以呈现复杂的可视化。开发的工具将是达到目的的一种手段：一种超出数据先入为主的新的，高度视觉上的方式，而是让数据自言自语。这种机制将使研究人员在视觉和迅速探索大规模数据集时，以创造性地和协作参与他们的专家直觉，以表明意外模式或异常值，以表明重要的发现。