A comparison of multi-label feature selection methods using the algorithm adaptation approach

Roiss Alhutaish, Nazlia Omar, Salwani Abdullah

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In a multi-label classification problem, each document is associated with a subset of labels. The documents often consist of multiple features. In addition, each document is usually associated with several labels. Therefore, feature selection is an important task in machine learning, which attempts to remove irrelevant and redundant features that can hinder the performance. This paper suggests transforming the multi-label documents into single-label documents before using the standard feature selection algorithm. Under this process, the document is copied into labels to which it belongs by adopting assigning all features to each label it belongs. With this context, we conducted a comparative study on five feature selection methods. These methods are incorporated into the traditional Naive Bayes classifiers, which are adapted to deal with multi-label documents. Experiments conducted with benchmark datasets showed that the multi-label Naive Bayes classifier coupled with the GSS method delivered a better performance than the MLNB classifier using other methods.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages199-212
Number of pages14
Volume9429
ISBN (Print)9783319259383, 9783319259383
DOIs
Publication statusPublished - 2015
Event4th International Visual Informatics Conference, IVIC 2015 - Bangi, Malaysia
Duration: 17 Nov 201519 Nov 2015

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9429
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other4th International Visual Informatics Conference, IVIC 2015
CountryMalaysia
CityBangi
Period17/11/1519/11/15

Fingerprint

Feature Selection
Feature extraction
Labels
Naive Bayes Classifier
Classifiers
Classification Problems
Comparative Study
Machine Learning
Classifier
Benchmark
Subset
Experiment
Learning systems

Keywords

  • Feature selection
  • Multi-label
  • Naive Bayes classifier

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Alhutaish, R., Omar, N., & Abdullah, S. (2015). A comparison of multi-label feature selection methods using the algorithm adaptation approach. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9429, pp. 199-212). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9429). Springer Verlag. https://doi.org/10.1007/978-3-319-25939-0_18

A comparison of multi-label feature selection methods using the algorithm adaptation approach. / Alhutaish, Roiss; Omar, Nazlia; Abdullah, Salwani.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9429 Springer Verlag, 2015. p. 199-212 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9429).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Alhutaish, R, Omar, N & Abdullah, S 2015, A comparison of multi-label feature selection methods using the algorithm adaptation approach. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 9429, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9429, Springer Verlag, pp. 199-212, 4th International Visual Informatics Conference, IVIC 2015, Bangi, Malaysia, 17/11/15. https://doi.org/10.1007/978-3-319-25939-0_18
Alhutaish R, Omar N, Abdullah S. A comparison of multi-label feature selection methods using the algorithm adaptation approach. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9429. Springer Verlag. 2015. p. 199-212. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-25939-0_18
Alhutaish, Roiss ; Omar, Nazlia ; Abdullah, Salwani. / A comparison of multi-label feature selection methods using the algorithm adaptation approach. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9429 Springer Verlag, 2015. pp. 199-212 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{afc329e07b8f47369e7b8099845cc82d,
title = "A comparison of multi-label feature selection methods using the algorithm adaptation approach",
abstract = "In a multi-label classification problem, each document is associated with a subset of labels. The documents often consist of multiple features. In addition, each document is usually associated with several labels. Therefore, feature selection is an important task in machine learning, which attempts to remove irrelevant and redundant features that can hinder the performance. This paper suggests transforming the multi-label documents into single-label documents before using the standard feature selection algorithm. Under this process, the document is copied into labels to which it belongs by adopting assigning all features to each label it belongs. With this context, we conducted a comparative study on five feature selection methods. These methods are incorporated into the traditional Naive Bayes classifiers, which are adapted to deal with multi-label documents. Experiments conducted with benchmark datasets showed that the multi-label Naive Bayes classifier coupled with the GSS method delivered a better performance than the MLNB classifier using other methods.",
keywords = "Feature selection, Multi-label, Naive Bayes classifier",
author = "Roiss Alhutaish and Nazlia Omar and Salwani Abdullah",
year = "2015",
doi = "10.1007/978-3-319-25939-0_18",
language = "English",
isbn = "9783319259383",
volume = "9429",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "199--212",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - A comparison of multi-label feature selection methods using the algorithm adaptation approach

AU - Alhutaish, Roiss

AU - Omar, Nazlia

AU - Abdullah, Salwani

PY - 2015

Y1 - 2015

N2 - In a multi-label classification problem, each document is associated with a subset of labels. The documents often consist of multiple features. In addition, each document is usually associated with several labels. Therefore, feature selection is an important task in machine learning, which attempts to remove irrelevant and redundant features that can hinder the performance. This paper suggests transforming the multi-label documents into single-label documents before using the standard feature selection algorithm. Under this process, the document is copied into labels to which it belongs by adopting assigning all features to each label it belongs. With this context, we conducted a comparative study on five feature selection methods. These methods are incorporated into the traditional Naive Bayes classifiers, which are adapted to deal with multi-label documents. Experiments conducted with benchmark datasets showed that the multi-label Naive Bayes classifier coupled with the GSS method delivered a better performance than the MLNB classifier using other methods.

AB - In a multi-label classification problem, each document is associated with a subset of labels. The documents often consist of multiple features. In addition, each document is usually associated with several labels. Therefore, feature selection is an important task in machine learning, which attempts to remove irrelevant and redundant features that can hinder the performance. This paper suggests transforming the multi-label documents into single-label documents before using the standard feature selection algorithm. Under this process, the document is copied into labels to which it belongs by adopting assigning all features to each label it belongs. With this context, we conducted a comparative study on five feature selection methods. These methods are incorporated into the traditional Naive Bayes classifiers, which are adapted to deal with multi-label documents. Experiments conducted with benchmark datasets showed that the multi-label Naive Bayes classifier coupled with the GSS method delivered a better performance than the MLNB classifier using other methods.

KW - Feature selection

KW - Multi-label

KW - Naive Bayes classifier

UR - http://www.scopus.com/inward/record.url?scp=84950160251&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84950160251&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-25939-0_18

DO - 10.1007/978-3-319-25939-0_18

M3 - Conference contribution

AN - SCOPUS:84950160251

SN - 9783319259383

SN - 9783319259383

VL - 9429

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 199

EP - 212

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -