Binarized Support Vector Machines

Emilio Carrizosa, Belén Martín-Barragán, Dolores Romero Morales

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

The widely used support vector machine (SVM) method has shown to yield very good results in supervised classification problems. Other methods such as classification trees have become more popular among practitioners than SVM thanks to their interpretability, which is an important issue in data mining.
In this work, we propose an SVM-based method that automatically detects the most important predictor variables and the role they play in the classifier. In particular, the proposed method is able to detect those values and intervals that are critical for the classification. The method involves the optimization of a linear programming problem in the spirit of the Lasso method with a large number of decision variables. The numerical experience reported shows that a rather direct use of the standard column generation strategy leads to a classification method that, in terms of classification ability, is competitive against the standard linear SVM and classification trees. Moreover, the proposed method is robust; i.e., it is stable in the presence of outliers and invariant to change of scale or measurement units of the predictor variables.
When the complexity of the classifier is an important issue, a wrapper feature selection method is applied, yielding simpler but still competitive classifiers.
The widely used support vector machine (SVM) method has shown to yield very good results in supervised classification problems. Other methods such as classification trees have become more popular among practitioners than SVM thanks to their interpretability, which is an important issue in data mining.
In this work, we propose an SVM-based method that automatically detects the most important predictor variables and the role they play in the classifier. In particular, the proposed method is able to detect those values and intervals that are critical for the classification. The method involves the optimization of a linear programming problem in the spirit of the Lasso method with a large number of decision variables. The numerical experience reported shows that a rather direct use of the standard column generation strategy leads to a classification method that, in terms of classification ability, is competitive against the standard linear SVM and classification trees. Moreover, the proposed method is robust; i.e., it is stable in the presence of outliers and invariant to change of scale or measurement units of the predictor variables.
When the complexity of the classifier is an important issue, a wrapper feature selection method is applied, yielding simpler but still competitive classifiers.
LanguageEnglish
JournalI N F O R M S Journal on Computing
Volume22
Issue number1
Pages154-167
ISSN1091-9856
DOIs
StatePublished - 2010
Externally publishedYes

Keywords

    Cite this

    Carrizosa, Emilio ; Martín-Barragán, Belén ; Morales, Dolores Romero. / Binarized Support Vector Machines. In: I N F O R M S Journal on Computing. 2010 ; Vol. 22, No. 1. pp. 154-167
    @article{4459e7e9567140bc90b5d7fa19286380,
    title = "Binarized Support Vector Machines",
    abstract = "The widely used support vector machine (SVM) method has shown to yield very good results in supervised classification problems. Other methods such as classification trees have become more popular among practitioners than SVM thanks to their interpretability, which is an important issue in data mining.In this work, we propose an SVM-based method that automatically detects the most important predictor variables and the role they play in the classifier. In particular, the proposed method is able to detect those values and intervals that are critical for the classification. The method involves the optimization of a linear programming problem in the spirit of the Lasso method with a large number of decision variables. The numerical experience reported shows that a rather direct use of the standard column generation strategy leads to a classification method that, in terms of classification ability, is competitive against the standard linear SVM and classification trees. Moreover, the proposed method is robust; i.e., it is stable in the presence of outliers and invariant to change of scale or measurement units of the predictor variables.When the complexity of the classifier is an important issue, a wrapper feature selection method is applied, yielding simpler but still competitive classifiers.",
    keywords = "Supervised classification, Binarization, Column generation, Support vector machines",
    author = "Emilio Carrizosa and Bel{\'e}n Mart{\'i}n-Barrag{\'a}n and Morales, {Dolores Romero}",
    year = "2010",
    doi = "10.1287/ijoc.1090.0317",
    language = "English",
    volume = "22",
    pages = "154--167",
    journal = "I N F O R M S Journal on Computing",
    issn = "1091-9856",
    publisher = "Institute for Operations Research and the Management Sciences",
    number = "1",

    }

    Carrizosa, E, Martín-Barragán, B & Morales, DR 2010, 'Binarized Support Vector Machines' I N F O R M S Journal on Computing, vol. 22, no. 1, pp. 154-167. DOI: 10.1287/ijoc.1090.0317

    Binarized Support Vector Machines. / Carrizosa, Emilio; Martín-Barragán, Belén; Morales, Dolores Romero.

    In: I N F O R M S Journal on Computing, Vol. 22, No. 1, 2010, p. 154-167.

    Research output: Contribution to journalJournal articleResearchpeer-review

    TY - JOUR

    T1 - Binarized Support Vector Machines

    AU - Carrizosa,Emilio

    AU - Martín-Barragán,Belén

    AU - Morales,Dolores Romero

    PY - 2010

    Y1 - 2010

    N2 - The widely used support vector machine (SVM) method has shown to yield very good results in supervised classification problems. Other methods such as classification trees have become more popular among practitioners than SVM thanks to their interpretability, which is an important issue in data mining.In this work, we propose an SVM-based method that automatically detects the most important predictor variables and the role they play in the classifier. In particular, the proposed method is able to detect those values and intervals that are critical for the classification. The method involves the optimization of a linear programming problem in the spirit of the Lasso method with a large number of decision variables. The numerical experience reported shows that a rather direct use of the standard column generation strategy leads to a classification method that, in terms of classification ability, is competitive against the standard linear SVM and classification trees. Moreover, the proposed method is robust; i.e., it is stable in the presence of outliers and invariant to change of scale or measurement units of the predictor variables.When the complexity of the classifier is an important issue, a wrapper feature selection method is applied, yielding simpler but still competitive classifiers.

    AB - The widely used support vector machine (SVM) method has shown to yield very good results in supervised classification problems. Other methods such as classification trees have become more popular among practitioners than SVM thanks to their interpretability, which is an important issue in data mining.In this work, we propose an SVM-based method that automatically detects the most important predictor variables and the role they play in the classifier. In particular, the proposed method is able to detect those values and intervals that are critical for the classification. The method involves the optimization of a linear programming problem in the spirit of the Lasso method with a large number of decision variables. The numerical experience reported shows that a rather direct use of the standard column generation strategy leads to a classification method that, in terms of classification ability, is competitive against the standard linear SVM and classification trees. Moreover, the proposed method is robust; i.e., it is stable in the presence of outliers and invariant to change of scale or measurement units of the predictor variables.When the complexity of the classifier is an important issue, a wrapper feature selection method is applied, yielding simpler but still competitive classifiers.

    KW - Supervised classification

    KW - Binarization

    KW - Column generation

    KW - Support vector machines

    U2 - 10.1287/ijoc.1090.0317

    DO - 10.1287/ijoc.1090.0317

    M3 - Journal article

    VL - 22

    SP - 154

    EP - 167

    JO - I N F O R M S Journal on Computing

    T2 - I N F O R M S Journal on Computing

    JF - I N F O R M S Journal on Computing

    SN - 1091-9856

    IS - 1

    ER -

    Carrizosa E, Martín-Barragán B, Morales DR. Binarized Support Vector Machines. I N F O R M S Journal on Computing. 2010;22(1):154-167. Available from, DOI: 10.1287/ijoc.1090.0317