Efficient computer experiment-based optimization through variable selection

Dachuan T. Shih, Seoung Bum Kim, Victoria C P Chen, Jay M. Rosenberger, Venkata L. Pilla

Research output: Contribution to journalArticle

9 Citations (Scopus)

Abstract

A computer experiment-based optimization approach employs design of experiments and statistical modeling to represent a complex objective function that can only be evaluated pointwise by running a computer model. In large-scale applications, the number of variables is huge, and direct use of computer experiments would require an exceedingly large experimental design and, consequently, significant computational effort. If a large portion of the variables have little impact on the objective, then there is a need to eliminate these before performing the complete set of computer experiments. This is a variable selection task. The ideal variable selection method for this task should handle unknown nonlinear structure, should be computationally fast, and would be conducted after a small number of computer experiment runs, likely fewer runs (N) than the number of variables (P). Conventional variable selection techniques are based on assumed linear model forms and cannot be applied in this "large P and small N" problem. In this paper, we present a framework that adds a variable selection step prior to computer experiment-based optimization, and we consider data mining methods, using principal components analysis and multiple testing based on false discovery rate, that are appropriate for our variable selection task. An airline fleet assignment case study is used to illustrate our approach.

Original languageEnglish
Pages (from-to)287-305
Number of pages19
JournalAnnals of Operations Research
Volume216
Issue number1
DOIs
Publication statusPublished - 2014 Jan 1

Fingerprint

Computer experiments
Variable selection
Assignment
Testing
Design of experiments
Experimental design
Principal component analysis
Modeling
Airlines
Objective function
Data mining

Keywords

  • Computer experiments
  • False discovery rate
  • Large-scale optimization
  • Regression trees
  • Variable selection

ASJC Scopus subject areas

  • Management Science and Operations Research
  • Decision Sciences(all)

Cite this

Efficient computer experiment-based optimization through variable selection. / Shih, Dachuan T.; Kim, Seoung Bum; Chen, Victoria C P; Rosenberger, Jay M.; Pilla, Venkata L.

In: Annals of Operations Research, Vol. 216, No. 1, 01.01.2014, p. 287-305.

Research output: Contribution to journalArticle

Shih, Dachuan T. ; Kim, Seoung Bum ; Chen, Victoria C P ; Rosenberger, Jay M. ; Pilla, Venkata L. / Efficient computer experiment-based optimization through variable selection. In: Annals of Operations Research. 2014 ; Vol. 216, No. 1. pp. 287-305.
@article{10799666b9d14b6297e10add2e35941e,
title = "Efficient computer experiment-based optimization through variable selection",
abstract = "A computer experiment-based optimization approach employs design of experiments and statistical modeling to represent a complex objective function that can only be evaluated pointwise by running a computer model. In large-scale applications, the number of variables is huge, and direct use of computer experiments would require an exceedingly large experimental design and, consequently, significant computational effort. If a large portion of the variables have little impact on the objective, then there is a need to eliminate these before performing the complete set of computer experiments. This is a variable selection task. The ideal variable selection method for this task should handle unknown nonlinear structure, should be computationally fast, and would be conducted after a small number of computer experiment runs, likely fewer runs (N) than the number of variables (P). Conventional variable selection techniques are based on assumed linear model forms and cannot be applied in this {"}large P and small N{"} problem. In this paper, we present a framework that adds a variable selection step prior to computer experiment-based optimization, and we consider data mining methods, using principal components analysis and multiple testing based on false discovery rate, that are appropriate for our variable selection task. An airline fleet assignment case study is used to illustrate our approach.",
keywords = "Computer experiments, False discovery rate, Large-scale optimization, Regression trees, Variable selection",
author = "Shih, {Dachuan T.} and Kim, {Seoung Bum} and Chen, {Victoria C P} and Rosenberger, {Jay M.} and Pilla, {Venkata L.}",
year = "2014",
month = "1",
day = "1",
doi = "10.1007/s10479-012-1129-y",
language = "English",
volume = "216",
pages = "287--305",
journal = "Annals of Operations Research",
issn = "0254-5330",
publisher = "Springer Netherlands",
number = "1",

}

TY - JOUR

T1 - Efficient computer experiment-based optimization through variable selection

AU - Shih, Dachuan T.

AU - Kim, Seoung Bum

AU - Chen, Victoria C P

AU - Rosenberger, Jay M.

AU - Pilla, Venkata L.

PY - 2014/1/1

Y1 - 2014/1/1

N2 - A computer experiment-based optimization approach employs design of experiments and statistical modeling to represent a complex objective function that can only be evaluated pointwise by running a computer model. In large-scale applications, the number of variables is huge, and direct use of computer experiments would require an exceedingly large experimental design and, consequently, significant computational effort. If a large portion of the variables have little impact on the objective, then there is a need to eliminate these before performing the complete set of computer experiments. This is a variable selection task. The ideal variable selection method for this task should handle unknown nonlinear structure, should be computationally fast, and would be conducted after a small number of computer experiment runs, likely fewer runs (N) than the number of variables (P). Conventional variable selection techniques are based on assumed linear model forms and cannot be applied in this "large P and small N" problem. In this paper, we present a framework that adds a variable selection step prior to computer experiment-based optimization, and we consider data mining methods, using principal components analysis and multiple testing based on false discovery rate, that are appropriate for our variable selection task. An airline fleet assignment case study is used to illustrate our approach.

AB - A computer experiment-based optimization approach employs design of experiments and statistical modeling to represent a complex objective function that can only be evaluated pointwise by running a computer model. In large-scale applications, the number of variables is huge, and direct use of computer experiments would require an exceedingly large experimental design and, consequently, significant computational effort. If a large portion of the variables have little impact on the objective, then there is a need to eliminate these before performing the complete set of computer experiments. This is a variable selection task. The ideal variable selection method for this task should handle unknown nonlinear structure, should be computationally fast, and would be conducted after a small number of computer experiment runs, likely fewer runs (N) than the number of variables (P). Conventional variable selection techniques are based on assumed linear model forms and cannot be applied in this "large P and small N" problem. In this paper, we present a framework that adds a variable selection step prior to computer experiment-based optimization, and we consider data mining methods, using principal components analysis and multiple testing based on false discovery rate, that are appropriate for our variable selection task. An airline fleet assignment case study is used to illustrate our approach.

KW - Computer experiments

KW - False discovery rate

KW - Large-scale optimization

KW - Regression trees

KW - Variable selection

UR - http://www.scopus.com/inward/record.url?scp=84897427654&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84897427654&partnerID=8YFLogxK

U2 - 10.1007/s10479-012-1129-y

DO - 10.1007/s10479-012-1129-y

M3 - Article

AN - SCOPUS:84897427654

VL - 216

SP - 287

EP - 305

JO - Annals of Operations Research

JF - Annals of Operations Research

SN - 0254-5330

IS - 1

ER -