Multivariate Analysis and Geovisualization with an Integrated Geographic Knowledge Discovery Approach

Area of Science:

• Geographic Information Science
• Data Visualization
• Computational Geography

Background:

• Understanding complex geographic problems requires effective methods for spatial pattern discovery.
• Multivariate spatial data presents challenges in interpretation and visualization.
• Existing methods may not fully integrate computational and visual analytical strengths.

Purpose of the Study:

• To develop and present an integrated approach for detecting and visualizing multivariate spatial patterns.
• To combine computational and visual methods for enhanced pattern discovery in large geographic datasets.
• To support interactive exploration and examination of complex spatial data.

Main Methods:

• Utilized the Self-Organizing Map (SOM) for multivariate, dimensional, and data reduction.
• Developed a systematic color scheme for encoding SOM results.
• Employed a modified Parallel Coordinate Plot (PCP) and a Geographic Map (GeoMap) for visualization.
• Incorporated human interaction for pattern exploration and examination.

Main Results:

• The integrated approach successfully detected and visualized multivariate spatial patterns.
• The Self-Organizing Map (SOM) effectively summarized large datasets into clusters.
• The combination of computational (SOM) and visual (PCP, GeoMap) methods proved effective.
• Mixed-initiative methods mitigated individual weaknesses, enabling efficient pattern discovery.

Conclusions:

• Integrated computational and visual methods offer a powerful framework for analyzing complex geographic problems.
• The developed approach facilitates effective and efficient discovery of multivariate spatial patterns.
• Interactive exploration enhances the interpretation and understanding of spatial data patterns.