Chen Jin, MacEachren Alan M, Guo Diansheng
Geo VISTA Center and Department of Geography, Pennsylvania State University, 302 Walker Building, University Park, PA16802,
Department of Geography, University of South Carolina, 709 Bull Street, Columbia, SC 29208,
Autocarto Res Symp. 2006 Jun;2006.
While many datasets carry geographic and temporal references, our ability to analyze these datasets lags behind our ability to collect them because of the challenges posed by both data complexity and scalability issues. This study develops a visual analytics approach that integrates human knowledge and judgments with visual, computational, and cartographic methods to support the application of visual analytics to relatively large spatio-temporal, multivariate datasets. Specifically, a variety of methods are employed for data clustering, pattern searching, information visualization and synthesis. By combining both human and machine strengths, this approach has a better chance to discover novel, relevant and potentially useful information that is difficult to detect by any method used in isolation. We demonstrate the effectiveness of the approach by applying the Visual Inquiry Toolkit we developed to analysis of a dataset containing geographically referenced, time-varying and multivariate data for U.S. technology industries.
虽然许多数据集都带有地理和时间参考信息，但由于数据复杂性和可扩展性问题带来的挑战，我们分析这些数据集的能力落后于收集它们的能力。本研究开发了一种可视化分析方法，该方法将人类知识和判断与视觉、计算和制图方法相结合，以支持将可视化分析应用于相对较大的时空多变量数据集。具体而言，采用了多种方法进行数据聚类、模式搜索、信息可视化和综合。通过结合人类和机器的优势，这种方法更有可能发现单独使用任何方法都难以检测到的新颖、相关且可能有用的信息。我们通过将我们开发的视觉探究工具包应用于对包含美国科技行业地理参考、随时间变化的多变量数据的数据集进行分析，来证明该方法的有效性。
