CAREER: A Visual Analysis Approach to Space-Time Data Exploration

从智能手机到健身追踪器，再到传感器建筑，数据目前正以前所未有的速度被收集。 现在，数据比以往任何时候都更能用来深入了解政策决策如何影响我们的日常生活。 例如，人们可以想象使用数据来帮助预测接下来可能发生犯罪的地方，或者为警察资源分配或饮食和活动模式的决策提供信息，可以用来提供改善个人整体健康和福祉的建议。 所有这些数据的基础是关于空间和时间的测量。 然而，在数据集中找到关系并准确地表示这些关系以通知政策变化是一个具有挑战性的问题。 这项研究解决了我们如何能够有效地探索这种时空数据，以加强知识发现和传播的基本问题。 这项研究既扩展了传统的视觉表征，又发展了新的视角，展示了相关性、聚类和其他各种空间动态随时间的变化。该研究计划的更广泛影响包括：（1）以新的视觉分析算法和开源软件的形式增强研究和教育基础设施;（2）在包括地理，城市规划和公共卫生在内的各个领域广泛传播视觉分析方法;（3）对社会的影响，包括传播改善公共卫生和安全的新工具和方法。 这个职业项目的主要教育目标是增加学生对关键但高度不可用的视觉分析技术的访问，并扩大对数据科学和工程的参与。为了实现这些目标，可视化分析教育计划将通过创新课程吸引广泛的学生群体（本科生和研究生），重点关注可视化数据分析和推动研究计划的核心技术（可视化分析工具）。通过关注这些技术及其在研究计划中的协同作用，教育计划直接将拟议的研究与教育相结合。这些项目将使当地和全球的多个群体（研究人员、患者、学生、代表性不足的群体）和机构（学术界、工业界、医疗保健、教育）受益。对于空间数据，将这些数据转换为可视化形式使用户能够快速查看模式、探索摘要并将表格形式中不明显的潜在地理现象的领域知识关联起来。然而，在可视化和探索这些大型时空数据集时，出现了一些关键的挑战。虽然，数据的基本地理组成部分很适合传统制图表示形式的单变量可视化（例如，等值线图、等值线图、不对称地图），随着数据变得多元，地图表示变得更加复杂。 多变量彩色地图、纹理、小倍数和3D视图已被用作在将空间数据绘制到地图时增加可传达的信息量的手段。 然而，这些方法中的每一种都有其自身的局限性。 多变量彩色地图和纹理导致认知过载，其中花费大量时间试图分离视觉通道中的数据元素。 在3D中，遮挡和杂乱仍然是有效视觉数据理解的基本挑战。利用小倍数可以帮助并排比较，但其可扩展性受到可用屏幕空间和与成对比较相关的认知开销的限制。 而不是局限于原来的时空域，这个建议旨在既扩展传统的视觉表示，并开发新的视图，显示如何相关性，集群和其他各种空间动态随时间的变化。 这些新观点的基础也是对视觉表示的需求，其中对表示的操作直接与底层的计算分析联系在一起。具体而言，本研究的重点是从城市规划，地理，公共卫生和犯罪的数据集，以解决：（1）提取半监督模板的空间和时间的聚合;（2）开发的交互技术，可视化转向和分类的时空数据;（3）集成多族异常检测算法和信息论方法进行半监督异常检测;（4）从时空数据中提取流场的新算法。更多信息可在项目网站（http：//vader.lab.asu.edu/Space-TimeVA）上查阅，包括开放源码软件、课程学习模块和播客。

From smart phones to fitness trackers to sensor enabled buildings, data is currently being collected at an unprecedented rate. Now, more than ever, data exists that can be used to gain insight into how policy decisions can impact our daily lives. For example, one can imagine using data to help predict where crime may occur next or inform decisions on police resource allocations or diet and activity patterns could be used to provide recommendations for improving an individual's overall health and well-being. Underlying all of this data are measurements with respect to space and time. However, finding relationships within datasets and accurately representing these relationships to inform policy changes is a challenging problem. This research addresses fundamental questions of how we can effectively explore such space-time data in order to enhance knowledge discovery and dissemination. This research both extends traditional visual representations and develops novel views for showing how correlations, clusters and other various spatial dynamics change over time. Broader impacts of the research program include: (1) enhanced infrastructure for research and education in the form of new visual analytics algorithms and open source software; (2) broad dissemination of visual analysis methods across various domains including geography, urban planning, and public health; and (3) impacts on society including the dissemination of novel tools and methods for improved public health and safety. The primary educational goals of this CAREER project are to increase students' access to crucial but highly unavailable visual analytic technologies and to broaden participation in data science and engineering. Toward those ends, the Visual Analytics Education program will engage broad student populations (undergraduate and graduate) through innovative curricula focusing on visual data analysis and the core technologies that drive the research program (visual analytics tools). By focusing on those technologies and their synergy in the research program, the education program directly integrates the proposed research with education. The programs will benefit multiple groups (researchers, patients, students, underrepresented groups) and institutions (academia, industry, healthcare, education) both locally and globally.For spatial data, the translation of such data into a visual form allows users to quickly see patterns, explore summaries and relate domain knowledge about underlying geographical phenomena that would not be apparent in tabular form. However, several critical challenges arise when visualizing and exploring these large spatiotemporal datasets. While, the underlying geographical component of the data lends itself well to univariate visualization in the form of traditional cartographic representations (e.g., choropleth, isopleth, dasymetric maps), as the data becomes multivariate, cartographic representations become more complex. Multivariate color maps, textures, small multiples and 3D views have been employed as means of increasing the amount of information that can be conveyed when plotting spatial data to a map. However, each of these methods has their own limitations. Multivariate color maps and textures result in cognitive overload where much time is spent trying to separate data elements in the visual channel. In 3D, occlusion and clutter remain fundamental challenges for effective visual data understanding. Utilizing small multiples can help in side-by-side comparison, but their scalability is limited by the available screen space and the cognitive overhead associated with pairwise comparisons. Instead of being confined to the original spatiotemporal domain, this proposal seeks to both extend traditional visual representations and develop novel views for showing how correlations, clusters and other various spatial dynamics change over time. Underlying these novel views is also the need for visual representations in which the manipulation of the representation is directly tied to the underlying computational analytics. Specifically, this research focuses on datasets from urban planning, geography, public health and crime to address: (1) the extraction of semi-supervised templates for spatial and temporal aggregation; (2) the development of interaction techniques for visual steering and classification of spatiotemporal data; (3) the integration of multiple families of anomaly detection algorithms and information theoretic methods for semi-supervised anomaly detection, and; (4) novel algorithms for the extraction of flow fields from spatiotemporal data. Additional information can be found at the project website (http://vader.lab.asu.edu/Space-TimeVA) including open source software, course learning modules and podcasts.