The Thrill of Discovery: Information Visualization as a Telescope for High Dimensions, by Ben Shneiderman

Originally published February 9, 2005

http://informationvisualization.typepad.com/sigvis/2005/02/the_thrill_of_d.html

Interactive information visualization provides researchers with remarkable tools for discovery. By combining powerful data mining methods with user-controlled interfaces, users are beginning to benefit from these potent telescopes for high-dimensional spaces. They can begin with an overview, zoom in on areas of interest, filter out unwanted items, and then click for details-on-demand. With careful design and efficient algorithms, the dynamic queries approach to data exploration can provide 100msec updates even for million-record databases.

Researchers respond by polishing their designs, conducting more realistic evaluations, and developing integrated solutions that start from data gathering and go to dissemination of insights. Developers know that they must tune their applications to specific professional domains such as gene expression analysis, financial data, or terror threat assessments. They are also looking for broader commercial applications such as ebay auctions, airline reservations, or consumer product shopping.

Our recent research includes:

1) TimeSearcher for visual exploration of large time series data in auctions, meteorology, and oil/gas discovery (www.cs.umd.edu/hcil/timesearcher).

2) Hierarchical Clustering Explorer 3.0 that now includes the rank-by-feature framework (www.cs.umd.edu/hcil/hce). By judiciously choosing from appropriate ranking criteria for low-dimensional axis-parallel projections, users can locate desired features of higher dimensional spaces.

3) Treemap 4.0 for exploring hierarchical data sets such as the gene ontology, digital libraries, public health data, and and portfolio management (www.cs.umd.edu/hcil/treemap).

The growing commercial success stories such as www.spotfire.com, www.smartmoney.com/marketmap and www.hivegroup.com reflect a gradually shifting popular acceptance of these novel approaches. Some users love these visualizations on first sight and respond with enthusiasm. Others take longer to grasp what they are seeing, and more importantly to realize how they might apply such tools to their data. They get hooked when they realize that information visualizations often present them with answers to questions that they didn’t even have: outliers stand out, trends become apparent, clusters make sense, and gaps invite attention.

Ben Shneiderman ben@cs.umd.edu
Founding Director, Human Computer Interaction Lab
Dept of Computer Science, University of Maryland
College Park, MD 20742 www.cs.umd.edu/~ben