Project Brief
General Competition (October 1997)Building a Future DatabaseDevelop software tools applying new techniques in machine learning and data mining to give business managers powerful predictive data models represented as common business databases that can be queried with familiar database tools. Sponsor: Astoria Software (formerly Links2Go.com Inc./Continum Software)800 West Cummings ParkSuite 4950 Woburn, MA 01801
Businesses keep records and databases because it's vitally important to know what has happened in the past. However, it's equally important to know what's going to happen in the future. That is why Continuum Software proposes to build databases of future events. Continuum's project meshes elements of "data mining"--the sifting of huge amounts of existing data such as sales records to pick up subtle patterns--and modern machine learning tools for predictive analysis. Several techniques exist to predict future trends based on data mining, but they generally require experts in artificial intelligence and machine learning. Continuum is developing software tools that use a new and unique application of machine learning called "Projective Visualization" to analyze huge amounts of warehoused data, even when the records in that warehouse are often incomplete, and predict the future behavior of not just individual variables but groups of variables: essentially predicting future database records. Business managers then can use their existing database tools--tools they are familiar with and that already have been tailored to their needs--to analyze these future records. The project involves several risky unknowns: for example, the amount of data needed for reasonable predictions may be too huge to process practically, and the kinds of data stored in typical business databases (e.g., transaction records) may be too limited for useful predictions. The project also will develop a graphical interface to simplify the development of the basic knowledge models used by the Projective Visualization technology. The value to U.S. industry of data mining has been demonstrated in the marketplace--one application in the credit-card industry saved 12 times the development costs in a single year. If successful, the Continuum technology will enable an entire class of powerful new data mining tools for market analysis and business simulations.
|
|
ATP website comments: webmaster-atp@nist.gov
Privacy Statement / Security Notice • NIST Disclaimer • NIST Information Quality Standards NIST is an agency of the U.S. Commerce Department |