|
Activity Number:
|
540
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics and Marketing
|
| Abstract - #305564 |
|
Title:
|
Hybrid GA-Based Constrained Optimization and Search for High-Potential Product Configurations
|
|
Author(s):
|
Brandon L. Paris*+ and Lynd D. Bacon and Jeff Hunter and Ashwin Sridhar
|
|
Companies:
|
General Mills, Inc. and Sighthound Solutions, Inc. and General Mills, Inc. and Sighthound Solutions, Inc.
|
|
Address:
|
One General Mills Blvd., Minneapolis, MN, 55426,
|
|
Keywords:
|
genetic algorithms ; fitness function ; optimization ; new product development
|
|
Abstract:
|
Manufacturers and service providers collect millions of product and concept evaluations every year. The data typically come from survey responses, and they undoubtedly contain valuable and unanticipated information about appealing new product ideas. Given the growing pressure on firms to innovate, it has become increasingly important to identify these ideas. We describe a system for searching for new product configurations using a hybrid genetic algorithm to perform constrained search and optimization. Our GA implementation is an extensible system that includes a configurable fitness function and a predictive modeling capability that allows it to "learn" parts of the fitness function with either parametric or nonparametric procedures. The GA is embedded in a multiuser environment that provides project management and collaboration functionality behind simple user interfaces.
|
