EXPLORATORY LEARNING FROM SPACE-ATTRIBUTE AGGREGATED DATA: A GEOVISUAL ANALYTICS APPROACH
Open Access
- Author:
- Chen, Jin
- Graduate Program:
- Geography
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- September 22, 2009
- Committee Members:
- Alan Maceachren, Dissertation Advisor/Co-Advisor
Alan Maceachren, Committee Chair/Co-Chair
Donna Jean Peuquet, Committee Member
Mark N Gahegan, Committee Member
Guoray Cai, Committee Member - Keywords:
- Spatial Epidemiology
Spatial Data Analysis
Geovisualization
Information Visualization
Geovisual Analytics
Spatial Data Mining - Abstract:
- The goal of this research is to develop a geovisual analytics framework and representative methods to support a scientific discovery process that addresses the following aspects of geographic inquiry: what, where and why. To address these aspects of geographic inquiry, the framework supports three groups of methods: spatial cluster analysis, multivariate data analysis, and spatial, multivariate analysis. As used here, these terms have broader meaning than the traditional ones in that they are conceptualized as supporting not only geographic data analysis, but also information analysis and knowledge discovery. Hence, the research develops, draws upon and combines new methods from multiple disciplines including (geo) information visualization, (geospatial) statistics, (geospatial) data mining, and cartography. The focus is on developing new visualization methods and integrating them with advanced data mining and statistical methods. The targeted data are geographically-referenced, high dimensional, areally-aggregated datasets of relatively large size (e.g., 3000+ U.S. counties, dozens of variables). The application domains to which the framework is directed primarily include spatial epidemiology and aspects of social and environmental science in which multivariate analysis of aggregated data is a central focus. The framework can be extended to support other domains in which geographic data analysis is relevant as well. The contributions of this research include: (1) considerably-enhanced methods for spatial cluster analysis, multivariate data analysis, and spatial-multivariate analysis, all supporting multi-scale analysis; (2) a geovisual-analytics prototype framework that combines visual, statistical and data mining methods together to support a complex exploratory learning process that integrates model-driven and data-driven methods together; (3) some advanced visualization methods including a dendrogram-matrix; (4) a software implementation of the geovisual analytics framework – the Visual Inquiry Toolkit (VIT); and (5) seven published papers.