Single factor analysis in MML mixture modelling

Russell T. Edwards; David L. Dowe

doi:10.1007/3-540-64383-4_9

Single factor analysis in MML mixture modelling

Russell T. Edwards, David L. Dowe

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

26 Citations (Scopus)

Abstract

Mixture modelling concerns the unsupervised discovery of clusters within data. Most current clustering algorithms assume that variables within classes are uncorrelated. We present a method for producing and evaluating models which account for inter-attribute correlation within classes with a single Gaussian linear factor. The method used is Minimum Message Length (MML), an invariant, information-theoretic Bayesian hypothesis evaluation criterion. Our work extends and unifies that of Wallace and Boulton (1968) and Wallace and Freeman (1992), concerned respectively with MML mixture modelling and MML single factor analysis. Results on simulated data are comparable to those of Wallace and Freeman (1992), outperforming Maximum Likelihood. We include an application of mixture modelling with single factors on spectral data from the Infrared Astronomical Satellite. Our model shows fewer unnecessary classes than that produced by AutoClass (Goebel et. al. 1989) due to the use of factors in modelling correlation.

Original language	English
Title of host publication	Research and Development in Knowledge Discovery and Data Mining - 2nd Pacific-Asia Conference, PAKDD 1998, Proceedings
Editors	Xindong Wu, Ramamohanarao Kotagiri, Kevin B. Korb
Publisher	Springer
Pages	96-109
Number of pages	14
ISBN (Print)	3540643834, 9783540643838
DOIs	https://doi.org/10.1007/3-540-64383-4_9
Publication status	Published - 1998
Event	Pacific-Asia Conference on Knowledge Discovery and Data Mining 1998 - Melbourne, Australia Duration: 15 Apr 1998 → 17 Apr 1998 Conference number: 2nd https://link.springer.com/book/10.1007/3-540-64383-4 (Proceedings)

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	1394
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	Pacific-Asia Conference on Knowledge Discovery and Data Mining 1998
Abbreviated title	PAKDD 1988
Country/Territory	Australia
City	Melbourne
Period	15/04/98 → 17/04/98
Internet address	https://link.springer.com/book/10.1007/3-540-64383-4 (Proceedings)

Keywords

Induction in KDD
Minimum message length
MML
Noise handling
Statistical and machine learning

Access to Document

10.1007/3-540-64383-4_9

Cite this

@inproceedings{7156cc33a0fd42599e325c71b510605d,

title = "Single factor analysis in MML mixture modelling",

abstract = "Mixture modelling concerns the unsupervised discovery of clusters within data. Most current clustering algorithms assume that variables within classes are uncorrelated. We present a method for producing and evaluating models which account for inter-attribute correlation within classes with a single Gaussian linear factor. The method used is Minimum Message Length (MML), an invariant, information-theoretic Bayesian hypothesis evaluation criterion. Our work extends and unifies that of Wallace and Boulton (1968) and Wallace and Freeman (1992), concerned respectively with MML mixture modelling and MML single factor analysis. Results on simulated data are comparable to those of Wallace and Freeman (1992), outperforming Maximum Likelihood. We include an application of mixture modelling with single factors on spectral data from the Infrared Astronomical Satellite. Our model shows fewer unnecessary classes than that produced by AutoClass (Goebel et. al. 1989) due to the use of factors in modelling correlation.",

keywords = "Induction in KDD, Minimum message length, MML, Noise handling, Statistical and machine learning",

author = "Edwards, {Russell T.} and Dowe, {David L.}",

year = "1998",

doi = "10.1007/3-540-64383-4_9",

language = "English",

isbn = "3540643834",

series = "Lecture Notes in Computer Science ",

publisher = "Springer",

pages = "96--109",

editor = "Xindong Wu and Ramamohanarao Kotagiri and Korb, {Kevin B.}",

booktitle = "Research and Development in Knowledge Discovery and Data Mining - 2nd Pacific-Asia Conference, PAKDD 1998, Proceedings",

note = "Pacific-Asia Conference on Knowledge Discovery and Data Mining 1998, PAKDD 1988 ; Conference date: 15-04-1998 Through 17-04-1998",

url = "https://link.springer.com/book/10.1007/3-540-64383-4",

}

Edwards, RT & Dowe, DL 1998, Single factor analysis in MML mixture modelling. in X Wu, R Kotagiri & KB Korb (eds), Research and Development in Knowledge Discovery and Data Mining - 2nd Pacific-Asia Conference, PAKDD 1998, Proceedings. Lecture Notes in Computer Science , vol. 1394, Springer, pp. 96-109, Pacific-Asia Conference on Knowledge Discovery and Data Mining 1998, Melbourne, Victoria, Australia, 15/04/98. https://doi.org/10.1007/3-540-64383-4_9

Single factor analysis in MML mixture modelling. / Edwards, Russell T.; Dowe, David L.
Research and Development in Knowledge Discovery and Data Mining - 2nd Pacific-Asia Conference, PAKDD 1998, Proceedings. ed. / Xindong Wu; Ramamohanarao Kotagiri; Kevin B. Korb. Springer, 1998. p. 96-109 (Lecture Notes in Computer Science ; Vol. 1394).

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

TY - GEN

T1 - Single factor analysis in MML mixture modelling

AU - Edwards, Russell T.

AU - Dowe, David L.

N1 - Conference code: 2nd

PY - 1998

Y1 - 1998

N2 - Mixture modelling concerns the unsupervised discovery of clusters within data. Most current clustering algorithms assume that variables within classes are uncorrelated. We present a method for producing and evaluating models which account for inter-attribute correlation within classes with a single Gaussian linear factor. The method used is Minimum Message Length (MML), an invariant, information-theoretic Bayesian hypothesis evaluation criterion. Our work extends and unifies that of Wallace and Boulton (1968) and Wallace and Freeman (1992), concerned respectively with MML mixture modelling and MML single factor analysis. Results on simulated data are comparable to those of Wallace and Freeman (1992), outperforming Maximum Likelihood. We include an application of mixture modelling with single factors on spectral data from the Infrared Astronomical Satellite. Our model shows fewer unnecessary classes than that produced by AutoClass (Goebel et. al. 1989) due to the use of factors in modelling correlation.

AB - Mixture modelling concerns the unsupervised discovery of clusters within data. Most current clustering algorithms assume that variables within classes are uncorrelated. We present a method for producing and evaluating models which account for inter-attribute correlation within classes with a single Gaussian linear factor. The method used is Minimum Message Length (MML), an invariant, information-theoretic Bayesian hypothesis evaluation criterion. Our work extends and unifies that of Wallace and Boulton (1968) and Wallace and Freeman (1992), concerned respectively with MML mixture modelling and MML single factor analysis. Results on simulated data are comparable to those of Wallace and Freeman (1992), outperforming Maximum Likelihood. We include an application of mixture modelling with single factors on spectral data from the Infrared Astronomical Satellite. Our model shows fewer unnecessary classes than that produced by AutoClass (Goebel et. al. 1989) due to the use of factors in modelling correlation.

KW - Induction in KDD

KW - Minimum message length

KW - MML

KW - Noise handling

KW - Statistical and machine learning

UR - http://www.scopus.com/inward/record.url?scp=77956962650&partnerID=8YFLogxK

U2 - 10.1007/3-540-64383-4_9

DO - 10.1007/3-540-64383-4_9

M3 - Conference Paper

AN - SCOPUS:77956962650

SN - 3540643834

SN - 9783540643838

T3 - Lecture Notes in Computer Science

SP - 96

EP - 109

BT - Research and Development in Knowledge Discovery and Data Mining - 2nd Pacific-Asia Conference, PAKDD 1998, Proceedings

A2 - Wu, Xindong

A2 - Kotagiri, Ramamohanarao

A2 - Korb, Kevin B.

PB - Springer

T2 - Pacific-Asia Conference on Knowledge Discovery and Data Mining 1998

Y2 - 15 April 1998 through 17 April 1998

ER -

Single factor analysis in MML mixture modelling

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Cite this