Clustering Categories in Support Vector Machines

Emilio Carrizosa, Amaya Nogales-Gómez, Dolores Romero Morales

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

60 Downloads (Pure)

Resumé

The support vector machine (SVM) is a state-of-the-art method in supervised classification. In this paper the Cluster Support Vector Machine (CLSVM) methodology is proposed with the aim to increase the sparsity of the SVM classifier in the presence of categorical features, leading to a gain in interpretability. The CLSVM methodology clusters categories and builds the SVM classifier in the clustered feature space. Four strategies for building the CLSVM classifier are presented based on solving: the SVM formulation in the original feature space, a quadratically constrained quadratic programming formulation, and a mixed integer quadratic programming formulation as well as its continuous relaxation. The computational study illustrates the performance of the CLSVM classifier using two clusters. In the tested datasets our methodology achieves comparable accuracy to that of the SVM in the original feature space, with a dramatic increase in sparsity.
OriginalsprogEngelsk
TidsskriftOmega
Vol/bind66
Udgave nummerPart A
Sider (fra-til)28-37
Antal sider20
ISSN0305-0483
DOI
StatusUdgivet - jan. 2017

Emneord

  • Support vector machine
  • Categorical features
  • Classifier sparsity
  • Clustering
  • Quadratically constrained programming
  • 0-1 programming

Citer dette

Carrizosa, Emilio ; Nogales-Gómez, Amaya ; Romero Morales, Dolores . / Clustering Categories in Support Vector Machines. I: Omega. 2017 ; Bind 66, Nr. Part A. s. 28-37.
@article{f57036cf7d44404ab999a3a638b5394c,
title = "Clustering Categories in Support Vector Machines",
abstract = "The support vector machine (SVM) is a state-of-the-art method in supervised classification. In this paper the Cluster Support Vector Machine (CLSVM) methodology is proposed with the aim to increase the sparsity of the SVM classifier in the presence of categorical features, leading to a gain in interpretability. The CLSVM methodology clusters categories and builds the SVM classifier in the clustered feature space. Four strategies for building the CLSVM classifier are presented based on solving: the SVM formulation in the original feature space, a quadratically constrained quadratic programming formulation, and a mixed integer quadratic programming formulation as well as its continuous relaxation. The computational study illustrates the performance of the CLSVM classifier using two clusters. In the tested datasets our methodology achieves comparable accuracy to that of the SVM in the original feature space, with a dramatic increase in sparsity.",
keywords = "Support vector machine, Categorical features, Classifier sparsity, Clustering, Quadratically constrained programming, 0-1 programming, Support vector machine, Categorical features, Classifier sparsity, Clustering, Quadratically constrained programming, 0-1 programming",
author = "Emilio Carrizosa and Amaya Nogales-G{\'o}mez and {Romero Morales}, Dolores",
year = "2017",
month = "1",
doi = "10.1016/j.omega.2016.01.008",
language = "English",
volume = "66",
pages = "28--37",
journal = "Omega: The International Journal of Management Science",
issn = "0305-0483",
publisher = "Elsevier",
number = "Part A",

}

Clustering Categories in Support Vector Machines. / Carrizosa, Emilio; Nogales-Gómez, Amaya; Romero Morales, Dolores .

I: Omega, Bind 66, Nr. Part A, 01.2017, s. 28-37.

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

TY - JOUR

T1 - Clustering Categories in Support Vector Machines

AU - Carrizosa, Emilio

AU - Nogales-Gómez, Amaya

AU - Romero Morales, Dolores

PY - 2017/1

Y1 - 2017/1

N2 - The support vector machine (SVM) is a state-of-the-art method in supervised classification. In this paper the Cluster Support Vector Machine (CLSVM) methodology is proposed with the aim to increase the sparsity of the SVM classifier in the presence of categorical features, leading to a gain in interpretability. The CLSVM methodology clusters categories and builds the SVM classifier in the clustered feature space. Four strategies for building the CLSVM classifier are presented based on solving: the SVM formulation in the original feature space, a quadratically constrained quadratic programming formulation, and a mixed integer quadratic programming formulation as well as its continuous relaxation. The computational study illustrates the performance of the CLSVM classifier using two clusters. In the tested datasets our methodology achieves comparable accuracy to that of the SVM in the original feature space, with a dramatic increase in sparsity.

AB - The support vector machine (SVM) is a state-of-the-art method in supervised classification. In this paper the Cluster Support Vector Machine (CLSVM) methodology is proposed with the aim to increase the sparsity of the SVM classifier in the presence of categorical features, leading to a gain in interpretability. The CLSVM methodology clusters categories and builds the SVM classifier in the clustered feature space. Four strategies for building the CLSVM classifier are presented based on solving: the SVM formulation in the original feature space, a quadratically constrained quadratic programming formulation, and a mixed integer quadratic programming formulation as well as its continuous relaxation. The computational study illustrates the performance of the CLSVM classifier using two clusters. In the tested datasets our methodology achieves comparable accuracy to that of the SVM in the original feature space, with a dramatic increase in sparsity.

KW - Support vector machine

KW - Categorical features

KW - Classifier sparsity

KW - Clustering

KW - Quadratically constrained programming

KW - 0-1 programming

KW - Support vector machine

KW - Categorical features

KW - Classifier sparsity

KW - Clustering

KW - Quadratically constrained programming

KW - 0-1 programming

UR - http://sfx-45cbs.hosted.exlibrisgroup.com/45cbs?url_ver=Z39.88-2004&url_ctx_fmt=info:ofi/fmt:kev:mtx:ctx&ctx_enc=info:ofi/enc:UTF-8&ctx_ver=Z39.88-2004&rfr_id=info:sid/sfxit.com:azlist&sfx.ignore_date_threshold=1&rft.object_id=954925430321&rft.object_portfolio_id=&svc.holdings=yes&svc.fulltext=yes

U2 - 10.1016/j.omega.2016.01.008

DO - 10.1016/j.omega.2016.01.008

M3 - Journal article

VL - 66

SP - 28

EP - 37

JO - Omega: The International Journal of Management Science

JF - Omega: The International Journal of Management Science

SN - 0305-0483

IS - Part A

ER -