Graduate Student Profiles
Jin Chen (jxc93[at] psu.edu) defended his doctoral thesis last semester, having worked under Dr. Alan MacEachren. During his time at the GeoVISTA Center, first as research staff and later as graduate student, Chen's scientific pursuits have had two foci: (1) developing methods for data analysis and knowledge construction from spatial, temporal, and multivariate data; and (2) applying those methods to address practical problems in sub-disciplines of geography including public health, environment, natural hazards, economics, and so on. His research adopts multidisciplinary approaches ranging from cartography to visualization, statistics, computation, data mining and artificial intelligence.
Jin's Thesis Work
Exploratory Learning from Space-Attribute Aggregated Data - A Geovisual Analytics Approach
Geographic inquiry, which has become increasingly important in many research domains, often involves three fundamental questions: what, where and why. For example, public health studies can pose the following questions: where are the places that have high mortality for a disease? What are the factors that have caused high mortality in those places (low-income, poor health services, etc.)? Similar questions can be posed in other social sciences (e.g., crime analysis): where are the high risk regions for a crime? Is the crime associated with certain socio-economic characteristics? Jin's thesis aimed to develop a geovisual analytics framework to support a scientific discovery process in geographic inquiry, utilizing three groups of methods: spatial cluster analysis, multivariate data analysis, and spatial, multivariate analysis. The research develops, draws upon and combines new methods from multiple disciplines including (geo) information visualization, (geospatial) statistics, (geospatial) data mining, and cartography. To demonstrate these new integrated methods, two geographically-referenced, high dimensional datasets of relatively large size - U.S. cervical cancer mortality rates by county from 2000-2004 and U.S. vehicle theft and population data by county for 2000 - were used as a proof of concept.
Recent papers resulting from Jin's thesis work:
- Chen, J., R. Roth, et al. (2008). "Geovisual analytics to enhance spatial scan statistic interpretation: an analysis of U.S. cervical cancer mortality." International Journal of Health Geographics 7(1): 57.
- Chen, J., A. M. MacEachren, et al. (2009). Constructing Overview + Detail Dendrogram-Matrix Views. IEEE Symposium on Information Visualization 2009 (INFOVIS 2009), Atlantic City, New Jersey USA. Associated Video Demonstration.
More about Jin's development of the VIT:
Building on the open-source Java framework of GeoVISTA Studio, Jin developed the Visual Inquiry Toolkit (VIT). The VIT provides a visual programming environment for spatial data analysis by integrating visual, computational, and cartographic methods, enabling human knowledge and judgment to be coupled productively with computational methods for incrementally searching patterns. The integrated VIT approach to geo-visual analytic methods and tools is able to:
- perform multivariate analysis (including time series analysis) with the Self-Organizing Map (SOM)
- encode the SOM result with colors derived from the Color-BrewerPlus component, which produces a 2D diverging-diverging color scheme
- visualize the data in a hierarchical data matrix view
- visualize the multivariate patterns with a modified Parallel Coordinate Plot (PCP) display and a map matrix
- support human interactions to explore and examine patterns
Jin initially developed the VIT for entry in the 2005 IEEE Information
Visualization Contest (www.infovis.org), where it took first place.
Contestants had to develop interesting and insightful ways of visualizing
and analyzing a complex benchmark data set provided by the contest
organizers. The 2005 data set contained geographically-referenced
statistics on employment, sales, and company relocations within selected
U.S. industries over a twelve year period. The GeoVISTA Center's winning
submission, led by Jin, demonstrated geo-visual analytic strategies
for detecting and exploring multivariate, spatio-temporal patterns.