Learning a Bayesian network classifier by jointly maximizing accuracy and information

Dan Halbersberg; Boaz Lerner

doi:10.3233/978-1-61499-672-9-1638

Learning a Bayesian network classifier by jointly maximizing accuracy and information

Dan Halbersberg, Boaz Lerner

Department of Industrial Engineering and Management

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

3 Scopus citations

Abstract

Although recent studies have shown that a Bayesian network classifier (BNC) that maximizes the classification accuracy (i.e., minimizes the 0/1 loss function) is a powerful tool in knowledge representation and classification, this classifier focuses on the majority class, is usually uninformative about the distribution of misclassifications, and is insensitive to error severity (making no distinction between misclassification types). We propose to learn a BNC using an information measure (IM) that jointly maximizes classification and information, and evaluate this measure using various databases. We show that an IM-based BNC is superior to BNCs learned using other measures, especially for ordinal classification and imbalanced problems, and does not fall behind state-of-the-art algorithms with respect to accuracy and amount of information provided.

Original language	English
Title of host publication	Frontiers in Artificial Intelligence and Applications
Editors	Gal A. Kaminka, Maria Fox, Paolo Bouquet, Eyke Hullermeier, Virginia Dignum, Frank Dignum, Frank van Harmelen
Publisher	IOS Press BV
Pages	1638-1639
Number of pages	2
Volume	285
ISBN (Electronic)	9781614996712
DOIs	https://doi.org/10.3233/978-1-61499-672-9-1638
State	Published - 1 Jan 2016
Event	22nd European Conference on Artificial Intelligence, ECAI 2016 - The Hague, Netherlands Duration: 29 Aug 2016 → 2 Sep 2016

Conference

Conference	22nd European Conference on Artificial Intelligence, ECAI 2016
Country/Territory	Netherlands
City	The Hague
Period	29/08/16 → 2/09/16

ASJC Scopus subject areas

Artificial Intelligence

Access to Document

10.3233/978-1-61499-672-9-1638

Cite this

@inproceedings{713897b4a97e41cd8d500952a9c8e50c,

title = "Learning a Bayesian network classifier by jointly maximizing accuracy and information",

abstract = "Although recent studies have shown that a Bayesian network classifier (BNC) that maximizes the classification accuracy (i.e., minimizes the 0/1 loss function) is a powerful tool in knowledge representation and classification, this classifier focuses on the majority class, is usually uninformative about the distribution of misclassifications, and is insensitive to error severity (making no distinction between misclassification types). We propose to learn a BNC using an information measure (IM) that jointly maximizes classification and information, and evaluate this measure using various databases. We show that an IM-based BNC is superior to BNCs learned using other measures, especially for ordinal classification and imbalanced problems, and does not fall behind state-of-the-art algorithms with respect to accuracy and amount of information provided.",

author = "Dan Halbersberg and Boaz Lerner",

note = "Frontiers in Artificial Intelligence and Applications ISSN (Print): 0922-6389; 22nd European Conference on Artificial Intelligence, ECAI 2016 ; Conference date: 29-08-2016 Through 02-09-2016",

year = "2016",

month = jan,

day = "1",

doi = "10.3233/978-1-61499-672-9-1638",

language = "English",

volume = "285",

pages = "1638--1639",

editor = "Kaminka, \{Gal A.\} and Maria Fox and Paolo Bouquet and Eyke Hullermeier and Virginia Dignum and Frank Dignum and \{van Harmelen\}, Frank",

booktitle = "Frontiers in Artificial Intelligence and Applications",

publisher = "IOS Press BV",

address = "Netherlands",

}

Halbersberg, D & Lerner, B 2016, Learning a Bayesian network classifier by jointly maximizing accuracy and information. in GA Kaminka, M Fox, P Bouquet, E Hullermeier, V Dignum, F Dignum & F van Harmelen (eds), Frontiers in Artificial Intelligence and Applications. vol. 285, IOS Press BV, pp. 1638-1639, 22nd European Conference on Artificial Intelligence, ECAI 2016, The Hague, Netherlands, 29/08/16. https://doi.org/10.3233/978-1-61499-672-9-1638

Learning a Bayesian network classifier by jointly maximizing accuracy and information. / Halbersberg, Dan; Lerner, Boaz.
Frontiers in Artificial Intelligence and Applications. ed. / Gal A. Kaminka; Maria Fox; Paolo Bouquet; Eyke Hullermeier; Virginia Dignum; Frank Dignum; Frank van Harmelen. Vol. 285 IOS Press BV, 2016. p. 1638-1639.

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

TY - GEN

T1 - Learning a Bayesian network classifier by jointly maximizing accuracy and information

AU - Halbersberg, Dan

AU - Lerner, Boaz

N1 - Frontiers in Artificial Intelligence and Applications ISSN (Print): 0922-6389

PY - 2016/1/1

Y1 - 2016/1/1

N2 - Although recent studies have shown that a Bayesian network classifier (BNC) that maximizes the classification accuracy (i.e., minimizes the 0/1 loss function) is a powerful tool in knowledge representation and classification, this classifier focuses on the majority class, is usually uninformative about the distribution of misclassifications, and is insensitive to error severity (making no distinction between misclassification types). We propose to learn a BNC using an information measure (IM) that jointly maximizes classification and information, and evaluate this measure using various databases. We show that an IM-based BNC is superior to BNCs learned using other measures, especially for ordinal classification and imbalanced problems, and does not fall behind state-of-the-art algorithms with respect to accuracy and amount of information provided.

AB - Although recent studies have shown that a Bayesian network classifier (BNC) that maximizes the classification accuracy (i.e., minimizes the 0/1 loss function) is a powerful tool in knowledge representation and classification, this classifier focuses on the majority class, is usually uninformative about the distribution of misclassifications, and is insensitive to error severity (making no distinction between misclassification types). We propose to learn a BNC using an information measure (IM) that jointly maximizes classification and information, and evaluate this measure using various databases. We show that an IM-based BNC is superior to BNCs learned using other measures, especially for ordinal classification and imbalanced problems, and does not fall behind state-of-the-art algorithms with respect to accuracy and amount of information provided.

UR - https://www.scopus.com/pages/publications/85013135311

U2 - 10.3233/978-1-61499-672-9-1638

DO - 10.3233/978-1-61499-672-9-1638

M3 - Conference contribution

AN - SCOPUS:85013135311

VL - 285

SP - 1638

EP - 1639

BT - Frontiers in Artificial Intelligence and Applications

A2 - Kaminka, Gal A.

A2 - Fox, Maria

A2 - Bouquet, Paolo

A2 - Hullermeier, Eyke

A2 - Dignum, Virginia

A2 - Dignum, Frank

A2 - van Harmelen, Frank

PB - IOS Press BV

T2 - 22nd European Conference on Artificial Intelligence, ECAI 2016

Y2 - 29 August 2016 through 2 September 2016

ER -

Learning a Bayesian network classifier by jointly maximizing accuracy and information

Abstract

Conference

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this