Individual attribute prior setting methods for nave Bayesian classifiers

Tzu-Tsung Wong, Liang Hao Chang

Research output: Contribution to journalArticle

11 Citations (Scopus)

Abstract

The generalized Dirichlet distribution has been shown to be a more appropriate prior for nave Bayesian classifiers, because it can release both the negative-correlation and the equal-confidence requirements of the Dirichlet distribution. The previous research did not take the impact of individual attributes on classification accuracy into account, and therefore assumed that all attributes follow the same generalized Dirichlet prior. In this study, the selective nave Bayes mechanism is employed to choose and rank attributes, and two methods are then proposed to search for the best prior of each single attribute according to the attribute ranks. The experimental results on 18 data sets show that the best approach is to use selective nave Bayes for filtering and ranking attributes when all of them have Dirichlet priors with Laplace's estimate. After the ranks of the chosen attributes are determined, individual setting is performed to search for the best noninformative generalized Dirichlet prior for each attribute. The selective nave Bayes is also compared with two representative filters for the feature selection, and the experimental results show that it has the best performance.

Original languageEnglish
Pages (from-to)1041-1047
Number of pages7
JournalPattern Recognition
Volume44
Issue number5
DOIs
Publication statusPublished - 2011 May 1

Fingerprint

Feature extraction
Classifiers

All Science Journal Classification (ASJC) codes

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Cite this

@article{627b1fec706a47d2baf2c50eeb97ba43,
title = "Individual attribute prior setting methods for nave Bayesian classifiers",
abstract = "The generalized Dirichlet distribution has been shown to be a more appropriate prior for nave Bayesian classifiers, because it can release both the negative-correlation and the equal-confidence requirements of the Dirichlet distribution. The previous research did not take the impact of individual attributes on classification accuracy into account, and therefore assumed that all attributes follow the same generalized Dirichlet prior. In this study, the selective nave Bayes mechanism is employed to choose and rank attributes, and two methods are then proposed to search for the best prior of each single attribute according to the attribute ranks. The experimental results on 18 data sets show that the best approach is to use selective nave Bayes for filtering and ranking attributes when all of them have Dirichlet priors with Laplace's estimate. After the ranks of the chosen attributes are determined, individual setting is performed to search for the best noninformative generalized Dirichlet prior for each attribute. The selective nave Bayes is also compared with two representative filters for the feature selection, and the experimental results show that it has the best performance.",
author = "Tzu-Tsung Wong and Chang, {Liang Hao}",
year = "2011",
month = "5",
day = "1",
doi = "10.1016/j.patcog.2010.11.002",
language = "English",
volume = "44",
pages = "1041--1047",
journal = "Pattern Recognition",
issn = "0031-3203",
publisher = "Elsevier Limited",
number = "5",

}

Individual attribute prior setting methods for nave Bayesian classifiers. / Wong, Tzu-Tsung; Chang, Liang Hao.

In: Pattern Recognition, Vol. 44, No. 5, 01.05.2011, p. 1041-1047.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Individual attribute prior setting methods for nave Bayesian classifiers

AU - Wong, Tzu-Tsung

AU - Chang, Liang Hao

PY - 2011/5/1

Y1 - 2011/5/1

N2 - The generalized Dirichlet distribution has been shown to be a more appropriate prior for nave Bayesian classifiers, because it can release both the negative-correlation and the equal-confidence requirements of the Dirichlet distribution. The previous research did not take the impact of individual attributes on classification accuracy into account, and therefore assumed that all attributes follow the same generalized Dirichlet prior. In this study, the selective nave Bayes mechanism is employed to choose and rank attributes, and two methods are then proposed to search for the best prior of each single attribute according to the attribute ranks. The experimental results on 18 data sets show that the best approach is to use selective nave Bayes for filtering and ranking attributes when all of them have Dirichlet priors with Laplace's estimate. After the ranks of the chosen attributes are determined, individual setting is performed to search for the best noninformative generalized Dirichlet prior for each attribute. The selective nave Bayes is also compared with two representative filters for the feature selection, and the experimental results show that it has the best performance.

AB - The generalized Dirichlet distribution has been shown to be a more appropriate prior for nave Bayesian classifiers, because it can release both the negative-correlation and the equal-confidence requirements of the Dirichlet distribution. The previous research did not take the impact of individual attributes on classification accuracy into account, and therefore assumed that all attributes follow the same generalized Dirichlet prior. In this study, the selective nave Bayes mechanism is employed to choose and rank attributes, and two methods are then proposed to search for the best prior of each single attribute according to the attribute ranks. The experimental results on 18 data sets show that the best approach is to use selective nave Bayes for filtering and ranking attributes when all of them have Dirichlet priors with Laplace's estimate. After the ranks of the chosen attributes are determined, individual setting is performed to search for the best noninformative generalized Dirichlet prior for each attribute. The selective nave Bayes is also compared with two representative filters for the feature selection, and the experimental results show that it has the best performance.

UR - http://www.scopus.com/inward/record.url?scp=78651346026&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78651346026&partnerID=8YFLogxK

U2 - 10.1016/j.patcog.2010.11.002

DO - 10.1016/j.patcog.2010.11.002

M3 - Article

VL - 44

SP - 1041

EP - 1047

JO - Pattern Recognition

JF - Pattern Recognition

SN - 0031-3203

IS - 5

ER -