A monotonic measure for optimal feature selection

Huan Liu; Hiroshi Motoda; Manoranjan Dash

A monotonic measure for optimal feature selection

Huan Liu, Hiroshi Motoda, Manoranjan Dash

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Feature selection is a problem of choosing a subset of relevant features. In general,only exhaustive search can bring about the optimal subset. With a monotonic measure, exhaustive search can be avoided without sacrificing optimality. Unfortunately, most error- or distance-based measures are not monotonic. A new measure is employed in this work that is monotonic and fast to compute. The search for relevant features according to this measure is guaranteed tobe complete but not exhaustive. Experiments are conducted for verification.

Original language	English (US)
Title of host publication	Machine Learning
Subtitle of host publication	ECML-1998 - 10th European Conference on Machine Learning, Proceedings
Editors	Claire Nédellec, Céline Rouveirol
Publisher	Springer Verlag
Pages	101-106
Number of pages	6
ISBN (Print)	3540644172, 9783540644170
State	Published - 1998
Externally published	Yes
Event	10th European Conference on Machine Learning, ECML 1998 - Chemnitz, Germany Duration: Apr 21 1998 → Apr 23 1998

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	1398

Other

Other	10th European Conference on Machine Learning, ECML 1998
Country/Territory	Germany
City	Chemnitz
Period	4/21/98 → 4/23/98

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Cite this

Liu, H., Motoda, H., & Dash, M. (1998). A monotonic measure for optimal feature selection. In C. Nédellec, & C. Rouveirol (Eds.), Machine Learning: ECML-1998 - 10th European Conference on Machine Learning, Proceedings (pp. 101-106). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1398). Springer Verlag.

A monotonic measure for optimal feature selection. / Liu, Huan; Motoda, Hiroshi; Dash, Manoranjan.
Machine Learning: ECML-1998 - 10th European Conference on Machine Learning, Proceedings. ed. / Claire Nédellec; Céline Rouveirol. Springer Verlag, 1998. p. 101-106 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1398).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Liu, H, Motoda, H & Dash, M 1998, A monotonic measure for optimal feature selection. in C Nédellec & C Rouveirol (eds), Machine Learning: ECML-1998 - 10th European Conference on Machine Learning, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1398, Springer Verlag, pp. 101-106, 10th European Conference on Machine Learning, ECML 1998, Chemnitz, Germany, 4/21/98.

Liu H, Motoda H, Dash M. A monotonic measure for optimal feature selection. In Nédellec C, Rouveirol C, editors, Machine Learning: ECML-1998 - 10th European Conference on Machine Learning, Proceedings. Springer Verlag. 1998. p. 101-106. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

Liu, Huan ; Motoda, Hiroshi ; Dash, Manoranjan. / A monotonic measure for optimal feature selection. Machine Learning: ECML-1998 - 10th European Conference on Machine Learning, Proceedings. editor / Claire Nédellec ; Céline Rouveirol. Springer Verlag, 1998. pp. 101-106 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{b90335e20f93447d8126db0563d5c28d,

title = "A monotonic measure for optimal feature selection",

abstract = "Feature selection is a problem of choosing a subset of relevant features. In general,only exhaustive search can bring about the optimal subset. With a monotonic measure, exhaustive search can be avoided without sacrificing optimality. Unfortunately, most error- or distance-based measures are not monotonic. A new measure is employed in this work that is monotonic and fast to compute. The search for relevant features according to this measure is guaranteed tobe complete but not exhaustive. Experiments are conducted for verification.",

author = "Huan Liu and Hiroshi Motoda and Manoranjan Dash",

note = "Publisher Copyright: {\textcopyright} Springer-Veriag Berlin Heidelberg 1998.; 10th European Conference on Machine Learning, ECML 1998 ; Conference date: 21-04-1998 Through 23-04-1998",

year = "1998",

language = "English (US)",

isbn = "3540644172",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "101--106",

editor = "Claire N{\'e}dellec and C{\'e}line Rouveirol",

booktitle = "Machine Learning",

}

TY - GEN

T1 - A monotonic measure for optimal feature selection

AU - Liu, Huan

AU - Motoda, Hiroshi

AU - Dash, Manoranjan

N1 - Publisher Copyright: © Springer-Veriag Berlin Heidelberg 1998.

PY - 1998

Y1 - 1998

N2 - Feature selection is a problem of choosing a subset of relevant features. In general,only exhaustive search can bring about the optimal subset. With a monotonic measure, exhaustive search can be avoided without sacrificing optimality. Unfortunately, most error- or distance-based measures are not monotonic. A new measure is employed in this work that is monotonic and fast to compute. The search for relevant features according to this measure is guaranteed tobe complete but not exhaustive. Experiments are conducted for verification.

AB - Feature selection is a problem of choosing a subset of relevant features. In general,only exhaustive search can bring about the optimal subset. With a monotonic measure, exhaustive search can be avoided without sacrificing optimality. Unfortunately, most error- or distance-based measures are not monotonic. A new measure is employed in this work that is monotonic and fast to compute. The search for relevant features according to this measure is guaranteed tobe complete but not exhaustive. Experiments are conducted for verification.

UR - http://www.scopus.com/inward/record.url?scp=79954993774&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79954993774&partnerID=8YFLogxK

M3 - Conference contribution

SN - 3540644172

SN - 9783540644170

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

SP - 101

EP - 106

BT - Machine Learning

A2 - Nédellec, Claire

A2 - Rouveirol, Céline

PB - Springer Verlag

T2 - 10th European Conference on Machine Learning, ECML 1998

Y2 - 21 April 1998 through 23 April 1998

ER -

A monotonic measure for optimal feature selection

Abstract

Publication series

Other

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this