Reduced multivariate polynomial-based neural network for automated traffic incident detection

D. Srinivasan, V. Sharma, K. A. Toh

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

This paper proposes a neural network model based on reduced multivariate polynomial pattern classifier for application in freeway incident detection. The reduced multivariate model (RM) is a recently proposed classifier model which is easy to implement and analyze, and has been observed to efficiently capture the nonlinear input-output relationships in many classification applications. Since the freeway incident detection can be treated as a two-category pattern classification problem, the reduced multivariate polynomial model is particularly suitable for this incident detection application. Both Recursive Singular Value Decomposition (RSVD)- based and gradient descent-based least square estimators were adopted to learn the RM classifier in this work. The comparison of results obtained with those from several other classification strategies demonstrates the efficacy of the proposed model for traffic incident detection.

Original languageEnglish
Pages (from-to)484-492
Number of pages9
JournalNeural Networks
Volume21
Issue number2-3
DOIs
Publication statusPublished - 2008 Mar 1

Fingerprint

Polynomials
Neural networks
Classifiers
Highway systems
Neural Networks (Computer)
Statistical Models
Least-Squares Analysis
Singular value decomposition
Pattern recognition

All Science Journal Classification (ASJC) codes

  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this

@article{0776c2cdaa764110aab9c4cd9ada1687,
title = "Reduced multivariate polynomial-based neural network for automated traffic incident detection",
abstract = "This paper proposes a neural network model based on reduced multivariate polynomial pattern classifier for application in freeway incident detection. The reduced multivariate model (RM) is a recently proposed classifier model which is easy to implement and analyze, and has been observed to efficiently capture the nonlinear input-output relationships in many classification applications. Since the freeway incident detection can be treated as a two-category pattern classification problem, the reduced multivariate polynomial model is particularly suitable for this incident detection application. Both Recursive Singular Value Decomposition (RSVD)- based and gradient descent-based least square estimators were adopted to learn the RM classifier in this work. The comparison of results obtained with those from several other classification strategies demonstrates the efficacy of the proposed model for traffic incident detection.",
author = "D. Srinivasan and V. Sharma and Toh, {K. A.}",
year = "2008",
month = "3",
day = "1",
doi = "10.1016/j.neunet.2007.12.028",
language = "English",
volume = "21",
pages = "484--492",
journal = "Neural Networks",
issn = "0893-6080",
publisher = "Elsevier Limited",
number = "2-3",

}

Reduced multivariate polynomial-based neural network for automated traffic incident detection. / Srinivasan, D.; Sharma, V.; Toh, K. A.

In: Neural Networks, Vol. 21, No. 2-3, 01.03.2008, p. 484-492.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Reduced multivariate polynomial-based neural network for automated traffic incident detection

AU - Srinivasan, D.

AU - Sharma, V.

AU - Toh, K. A.

PY - 2008/3/1

Y1 - 2008/3/1

N2 - This paper proposes a neural network model based on reduced multivariate polynomial pattern classifier for application in freeway incident detection. The reduced multivariate model (RM) is a recently proposed classifier model which is easy to implement and analyze, and has been observed to efficiently capture the nonlinear input-output relationships in many classification applications. Since the freeway incident detection can be treated as a two-category pattern classification problem, the reduced multivariate polynomial model is particularly suitable for this incident detection application. Both Recursive Singular Value Decomposition (RSVD)- based and gradient descent-based least square estimators were adopted to learn the RM classifier in this work. The comparison of results obtained with those from several other classification strategies demonstrates the efficacy of the proposed model for traffic incident detection.

AB - This paper proposes a neural network model based on reduced multivariate polynomial pattern classifier for application in freeway incident detection. The reduced multivariate model (RM) is a recently proposed classifier model which is easy to implement and analyze, and has been observed to efficiently capture the nonlinear input-output relationships in many classification applications. Since the freeway incident detection can be treated as a two-category pattern classification problem, the reduced multivariate polynomial model is particularly suitable for this incident detection application. Both Recursive Singular Value Decomposition (RSVD)- based and gradient descent-based least square estimators were adopted to learn the RM classifier in this work. The comparison of results obtained with those from several other classification strategies demonstrates the efficacy of the proposed model for traffic incident detection.

UR - http://www.scopus.com/inward/record.url?scp=40649118931&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=40649118931&partnerID=8YFLogxK

U2 - 10.1016/j.neunet.2007.12.028

DO - 10.1016/j.neunet.2007.12.028

M3 - Article

C2 - 18276107

AN - SCOPUS:40649118931

VL - 21

SP - 484

EP - 492

JO - Neural Networks

JF - Neural Networks

SN - 0893-6080

IS - 2-3

ER -