Determining the need for multi-label classifiers by measuring unexplained covariance

Laurence A.F. Park; Jesse Read

doi:10.1007/978-981-96-8295-9_18

Determining the need for multi-label classifiers by measuring unexplained covariance

Laurence A.F. Park
, Jesse Read

School of Computer, Data & Mathematical Sciences

Ecole Polytechnique

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

Abstract

Multi-label classifiers make use of associations between labels in multi-label data to increase the accuracy of prediction. Before using a multi-label classifier, the data should be analysed to identify if there are associations between labels. If sets of independent labels are found, the data can be split in to multiple smaller data sets for analysis. Unfortunately, each label is dependent on the set of observations and so measuring label dependence is futile. What we actually seek is independence after taking the observations into account. In this article, we examine the concepts of explained and unexplained label covariance for measuring label dependence. We explore the use of a Normal copula model for modelling the label dependence/covariance and show that it is not able to measure conditional covariance directly. We then propose a new statistical model that allows direct measurement of label covariance (both constant and conditional). The model is validated using generated data and it is also used to examine the label covariance in real world data, allowing us to build simpler multi-label models.

Original language	English
Title of host publication	Data Science: Foundations and Applications: 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, June 10-13, 2025, Proceedings, Part VI
Editors	Xintao Wu, Myra Spiliopoulou, Can Wang, Vipin Kumar, Longbing Cao, Xiangmin Zhou, Guansong Pang, Joao Gama
Place of Publication	Singapore
Publisher	Springer
Pages	250-261
Number of pages	12
ISBN (Electronic)	9789819682959
ISBN (Print)	9789819682942
DOIs	https://doi.org/10.1007/978-981-96-8295-9_18
Publication status	Published - 2025
Event	29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025 - Sydney, Australia Duration: 10 Jun 2025 → 13 Jun 2025

Publication series

Name	Lecture Notes in Computer Science
Volume	15875
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025
Country/Territory	Australia
City	Sydney
Period	10/06/25 → 13/06/25

Access to Document

10.1007/978-981-96-8295-9_18

Cite this

Park, L. A. F., & Read, J. (2025). Determining the need for multi-label classifiers by measuring unexplained covariance. In X. Wu, M. Spiliopoulou, C. Wang, V. Kumar, L. Cao, X. Zhou, G. Pang, & J. Gama (Eds.), Data Science: Foundations and Applications: 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, June 10-13, 2025, Proceedings, Part VI (pp. 250-261). (Lecture Notes in Computer Science; Vol. 15875). Springer. https://doi.org/10.1007/978-981-96-8295-9_18

Park, Laurence A.F. ; Read, Jesse. / Determining the need for multi-label classifiers by measuring unexplained covariance. Data Science: Foundations and Applications: 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, June 10-13, 2025, Proceedings, Part VI. editor / Xintao Wu ; Myra Spiliopoulou ; Can Wang ; Vipin Kumar ; Longbing Cao ; Xiangmin Zhou ; Guansong Pang ; Joao Gama. Singapore : Springer, 2025. pp. 250-261 (Lecture Notes in Computer Science).

@inbook{9f3f29d3653c48cab4a4503618acb6d0,

title = "Determining the need for multi-label classifiers by measuring unexplained covariance",

abstract = "Multi-label classifiers make use of associations between labels in multi-label data to increase the accuracy of prediction. Before using a multi-label classifier, the data should be analysed to identify if there are associations between labels. If sets of independent labels are found, the data can be split in to multiple smaller data sets for analysis. Unfortunately, each label is dependent on the set of observations and so measuring label dependence is futile. What we actually seek is independence after taking the observations into account. In this article, we examine the concepts of explained and unexplained label covariance for measuring label dependence. We explore the use of a Normal copula model for modelling the label dependence/covariance and show that it is not able to measure conditional covariance directly. We then propose a new statistical model that allows direct measurement of label covariance (both constant and conditional). The model is validated using generated data and it is also used to examine the label covariance in real world data, allowing us to build simpler multi-label models.",

author = "Park, \{Laurence A.F.\} and Jesse Read",

year = "2025",

doi = "10.1007/978-981-96-8295-9\_18",

language = "English",

isbn = "9789819682942",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "250--261",

editor = "Xintao Wu and Myra Spiliopoulou and Can Wang and Vipin Kumar and Longbing Cao and Xiangmin Zhou and Guansong Pang and Joao Gama",

booktitle = "Data Science: Foundations and Applications: 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, June 10-13, 2025, Proceedings, Part VI",

note = "29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025 ; Conference date: 10-06-2025 Through 13-06-2025",

}

Park, LAF & Read, J 2025, Determining the need for multi-label classifiers by measuring unexplained covariance. in X Wu, M Spiliopoulou, C Wang, V Kumar, L Cao, X Zhou, G Pang & J Gama (eds), Data Science: Foundations and Applications: 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, June 10-13, 2025, Proceedings, Part VI. Lecture Notes in Computer Science, vol. 15875, Springer, Singapore, pp. 250-261, 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, 10/06/25. https://doi.org/10.1007/978-981-96-8295-9_18

Determining the need for multi-label classifiers by measuring unexplained covariance. / Park, Laurence A.F.; Read, Jesse.
Data Science: Foundations and Applications: 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, June 10-13, 2025, Proceedings, Part VI. ed. / Xintao Wu; Myra Spiliopoulou; Can Wang; Vipin Kumar; Longbing Cao; Xiangmin Zhou; Guansong Pang; Joao Gama. Singapore: Springer, 2025. p. 250-261 (Lecture Notes in Computer Science; Vol. 15875).

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

TY - CHAP

T1 - Determining the need for multi-label classifiers by measuring unexplained covariance

AU - Park, Laurence A.F.

AU - Read, Jesse

PY - 2025

Y1 - 2025

N2 - Multi-label classifiers make use of associations between labels in multi-label data to increase the accuracy of prediction. Before using a multi-label classifier, the data should be analysed to identify if there are associations between labels. If sets of independent labels are found, the data can be split in to multiple smaller data sets for analysis. Unfortunately, each label is dependent on the set of observations and so measuring label dependence is futile. What we actually seek is independence after taking the observations into account. In this article, we examine the concepts of explained and unexplained label covariance for measuring label dependence. We explore the use of a Normal copula model for modelling the label dependence/covariance and show that it is not able to measure conditional covariance directly. We then propose a new statistical model that allows direct measurement of label covariance (both constant and conditional). The model is validated using generated data and it is also used to examine the label covariance in real world data, allowing us to build simpler multi-label models.

AB - Multi-label classifiers make use of associations between labels in multi-label data to increase the accuracy of prediction. Before using a multi-label classifier, the data should be analysed to identify if there are associations between labels. If sets of independent labels are found, the data can be split in to multiple smaller data sets for analysis. Unfortunately, each label is dependent on the set of observations and so measuring label dependence is futile. What we actually seek is independence after taking the observations into account. In this article, we examine the concepts of explained and unexplained label covariance for measuring label dependence. We explore the use of a Normal copula model for modelling the label dependence/covariance and show that it is not able to measure conditional covariance directly. We then propose a new statistical model that allows direct measurement of label covariance (both constant and conditional). The model is validated using generated data and it is also used to examine the label covariance in real world data, allowing us to build simpler multi-label models.

UR - https://www.scopus.com/pages/publications/105009792894

UR - https://go.openathens.net/redirector/westernsydney.edu.au?url=https://doi.org/10.1007/978-981-96-8295-9_18

U2 - 10.1007/978-981-96-8295-9_18

DO - 10.1007/978-981-96-8295-9_18

M3 - Chapter

AN - SCOPUS:105009792894

SN - 9789819682942

T3 - Lecture Notes in Computer Science

SP - 250

EP - 261

BT - Data Science: Foundations and Applications: 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, June 10-13, 2025, Proceedings, Part VI

A2 - Wu, Xintao

A2 - Spiliopoulou, Myra

A2 - Wang, Can

A2 - Kumar, Vipin

A2 - Cao, Longbing

A2 - Zhou, Xiangmin

A2 - Pang, Guansong

A2 - Gama, Joao

PB - Springer

CY - Singapore

T2 - 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025

Y2 - 10 June 2025 through 13 June 2025

ER -

Park LAF, Read J. Determining the need for multi-label classifiers by measuring unexplained covariance. In Wu X, Spiliopoulou M, Wang C, Kumar V, Cao L, Zhou X, Pang G, Gama J, editors, Data Science: Foundations and Applications: 29th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2025, Sydney, Australia, June 10-13, 2025, Proceedings, Part VI. Singapore: Springer. 2025. p. 250-261. (Lecture Notes in Computer Science). doi: 10.1007/978-981-96-8295-9_18