Determining the saliency of input variables in neural network classifiers

Ravinder Nath, Balaji Rajagopalan, Randy Ryker

Research output: Contribution to journalArticle

74 Citations (Scopus)

Abstract

This paper examines a measure of the saliency of the input variables that is based upon the connection weights of the neural network. Using Monte Carlo simulation techniques, a comparison of this method with the traditional stepwise variable selection rule in Fisher's linear classification analysis (FLDA) is made. It is found that the method works quite well in identifying significant variables under a variety of experimental conditions, including neural network architectures and data configurations. In addition, data from acquired and liquidated firms is used to illustrate and validate the technique.

Original languageEnglish
Pages (from-to)767-773
Number of pages7
JournalComputers and Operations Research
Volume24
Issue number8
StatePublished - Aug 1997
Externally publishedYes

Fingerprint

Saliency
neural network
Classifiers
Classifier
Neural Networks
Neural networks
comparison of methods
Selection Rules
Network Architecture
Variable Selection
Network architecture
Monte Carlo Simulation
firm
simulation
Configuration
Business
Monte Carlo simulation
Variable selection

All Science Journal Classification (ASJC) codes

  • Information Systems and Management
  • Management Science and Operations Research
  • Applied Mathematics
  • Modeling and Simulation
  • Transportation

Cite this

Determining the saliency of input variables in neural network classifiers. / Nath, Ravinder; Rajagopalan, Balaji; Ryker, Randy.

In: Computers and Operations Research, Vol. 24, No. 8, 08.1997, p. 767-773.

Research output: Contribution to journalArticle

Nath, Ravinder ; Rajagopalan, Balaji ; Ryker, Randy. / Determining the saliency of input variables in neural network classifiers. In: Computers and Operations Research. 1997 ; Vol. 24, No. 8. pp. 767-773.
@article{25a2c828e44343b8aee1926811ccfb75,
title = "Determining the saliency of input variables in neural network classifiers",
abstract = "This paper examines a measure of the saliency of the input variables that is based upon the connection weights of the neural network. Using Monte Carlo simulation techniques, a comparison of this method with the traditional stepwise variable selection rule in Fisher's linear classification analysis (FLDA) is made. It is found that the method works quite well in identifying significant variables under a variety of experimental conditions, including neural network architectures and data configurations. In addition, data from acquired and liquidated firms is used to illustrate and validate the technique.",
author = "Ravinder Nath and Balaji Rajagopalan and Randy Ryker",
year = "1997",
month = "8",
language = "English",
volume = "24",
pages = "767--773",
journal = "Surveys in Operations Research and Management Science",
issn = "0305-0548",
publisher = "Elsevier Limited",
number = "8",

}

TY - JOUR

T1 - Determining the saliency of input variables in neural network classifiers

AU - Nath, Ravinder

AU - Rajagopalan, Balaji

AU - Ryker, Randy

PY - 1997/8

Y1 - 1997/8

N2 - This paper examines a measure of the saliency of the input variables that is based upon the connection weights of the neural network. Using Monte Carlo simulation techniques, a comparison of this method with the traditional stepwise variable selection rule in Fisher's linear classification analysis (FLDA) is made. It is found that the method works quite well in identifying significant variables under a variety of experimental conditions, including neural network architectures and data configurations. In addition, data from acquired and liquidated firms is used to illustrate and validate the technique.

AB - This paper examines a measure of the saliency of the input variables that is based upon the connection weights of the neural network. Using Monte Carlo simulation techniques, a comparison of this method with the traditional stepwise variable selection rule in Fisher's linear classification analysis (FLDA) is made. It is found that the method works quite well in identifying significant variables under a variety of experimental conditions, including neural network architectures and data configurations. In addition, data from acquired and liquidated firms is used to illustrate and validate the technique.

UR - http://www.scopus.com/inward/record.url?scp=0031209593&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0031209593&partnerID=8YFLogxK

M3 - Article

VL - 24

SP - 767

EP - 773

JO - Surveys in Operations Research and Management Science

JF - Surveys in Operations Research and Management Science

SN - 0305-0548

IS - 8

ER -