|
Activity Number:
|
242
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #306644 |
|
Title:
|
InfoEvolve(TM): a Data Mining Tool Combining Information Theory and Genetic Algorithms
|
|
Author(s):
|
Aaron J. Owens*+ and Karen M. Bloch
|
|
Companies:
|
DuPont and DuPont
|
|
Address:
|
P.O. Box 80249, Wilmington, DE, 19880-0249,
|
|
Keywords:
|
information theory ; genetic algorithms ; data mining ; dimensionality reduction ; classification ; empirical modeling
|
|
Abstract:
|
We introduce the InfoEvolve(TM) empirical modeling technology, which is based on the combination of Information Theory and Genetic Algorithms. The InfoEvolve(TM) Discovery module is used for dimensionality reduction in large data sets, and its Predict module uses a committee of experts approach for accurate prediction of classifications. We present two example applications: using dimensionality reduction to improve model accuracy for a grain classification instrument, and modeling to predict movements of the S&P 500 stock index on a day-to-day basis.
|
