Issue No.05 - September/October (2004 vol.24)
Daniel A. Keim , University of Constance, Germany
Christian Panse , University of Constance, Germany
Mike Sips , University of Constance, Germany
Stephen C. North , AT&T Labs
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MCG.2004.41
The information revolution is creating and publishing vast data sets, such as records of business transactions, environmental statistics, and census demographics. In human versus application domains, this data is collected and indexed by geospatial location. The discovery of interesting patterns in such databases through spatial data mining is a key to turning this raw data into valuable information. Challenges arise because newly available geospatial data sets often have millions of records, or even more. New techniques are needed to cope with this scale. The Wide Area Layout Data Observer (Waldo) is a novel visual data mining system, based on PixelMaps, for analyzing large geospatial data sets. PixelMaps combine density-based distortion of map regions with local pixel repositioning to highlight clusters and avoid data loss from over plotting. To enhance data exploration, Waldo involves the human in cluster discovery.
Daniel A. Keim, Christian Panse, Mike Sips, Stephen C. North, "Visual Data Mining in Large Geospatial Point Sets", IEEE Computer Graphics and Applications, vol.24, no. 5, pp. 36-44, September/October 2004, doi:10.1109/MCG.2004.41