|
Activity Number:
|
65
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #306398 |
|
Title:
|
An Application of Genetic Algorithms to Multivariate Optimal Allocation in Stratified Sample Designs
|
|
Author(s):
|
Charles Day*+
|
|
Companies:
|
U.S. Internal Revenue Service
|
|
Address:
|
P.O. Box 2608, Washington, DC, 20013,
|
|
Keywords:
|
evolutionary algorithm ; stochastic search
|
|
Abstract:
|
Sampling statisticians are often interested in allocating sample units to strata in a stratified probability sample in such a way that variance constraints for two or more variables of interest are satisfied with least cost. This problem falls into the class of convex mathematical programming problems. It is usually solved numerically, using a program that searches for an arbitrarily close approximation to the optimum. Genetic Algorithms (GA's) use a model of computation based on biological evolution to perform a stochastic search of the solution space of an optimization problem. This paper reports on the use of a GA to solve the multivariate optimal allocation problem. Because of the flexibility of fitness functions (objective functions) in GA's, this approach has the potential to be extended to meet additional optimality criteria involving more complicated objectives.
|
