MetaLIRS: Meta-learning for Imputation and Regression Selection

Işıl Baysal Erez; Jan Flokstra; Mannes Poel; Maurice van Keulen

doi:10.1007/978-3-031-77731-8_15

MetaLIRS: Meta-learning for Imputation and Regression Selection

Işıl Baysal Erez^*
, Jan Flokstra
, Mannes Poel
, Maurice van Keulen

^*Corresponding author for this work

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

1 Citation (Scopus)

55 Downloads (Pure)

Abstract

Missing data is a prevalent problem in data science for many fields such as natural, social, and health sciences. Since most regression methods can not handle missing data directly, imputation methods are used in data pre-processing. Finding the best imputation method is non-trivial, however. Moreover, our results show that an independent choice for a best imputation method does not always result in the best predictive performance in the end; the combination matters. Furthermore, search-based approaches for finding a best-fitting imputer/regressor-pair
can be computationally intensive. In this paper, we propose the MetaLIRS (Meta Learning Imputation and Regression Selection) frame-work for developing resource-friendly ML-based recommendation models for method selection. With MetaLIRS, we constructed a proof-of-concept
recommendation model based on 12 meta-features that achieves an accuracy of 63% for selecting the best-fitting imputer/regressor-pair. A data scientist can use this model for a quick resource-friendly recommendation on which imputation and regression method to use for their particular
data set and task without the need for an expensive grid search among methods.

Original language	English
Title of host publication	Intelligent Data Engineering and Automated Learning - IDEAL 2024
Subtitle of host publication	25th International Conference, Valencia, Spain, November 20-22, 2024. Proceedings, Part I
Editors	Vincente Julian, David Camacho, Hujun Yin, Juan M. Alberola, Vitor Beires Nogueira, Paulo Novais, Antonio Tallón-Ballesteros
Place of Publication	Cham, Switzerland
Publisher	Springer
Pages	155-166
Number of pages	12
ISBN (Electronic)	978-3-031-77731-8
ISBN (Print)	978-3-031-77730-1
DOIs	https://doi.org/10.1007/978-3-031-77731-8_15
Publication status	Published - 14 Nov 2024
Event	25th International Conference on Intelligent Data Engineering and Automated Learning - IDEAL 2024 - Valencia, Spain Duration: 20 Nov 2024 → 22 Nov 2024 Conference number: 25

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	15346
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	25th International Conference on Intelligent Data Engineering and Automated Learning - IDEAL 2024
Abbreviated title	IDEAL 2024
Country/Territory	Spain
City	Valencia
Period	20/11/24 → 22/11/24

Keywords

2024 OA procedure

Access to Document

10.1007/978-3-031-77731-8_15

978-3-031-77731-8_15Final published version, 883 KBLicence: Taverne

Cite this

Baysal Erez, I., Flokstra, J., Poel, M., & van Keulen, M. (2024). MetaLIRS: Meta-learning for Imputation and Regression Selection. In V. Julian, D. Camacho, H. Yin, J. M. Alberola, V. Beires Nogueira, P. Novais, & A. Tallón-Ballesteros (Eds.), Intelligent Data Engineering and Automated Learning - IDEAL 2024: 25th International Conference, Valencia, Spain, November 20-22, 2024. Proceedings, Part I (pp. 155-166). (Lecture Notes in Computer Science; Vol. 15346). Springer. https://doi.org/10.1007/978-3-031-77731-8_15

Baysal Erez, Işıl ; Flokstra, Jan ; Poel, Mannes et al. / MetaLIRS : Meta-learning for Imputation and Regression Selection. Intelligent Data Engineering and Automated Learning - IDEAL 2024: 25th International Conference, Valencia, Spain, November 20-22, 2024. Proceedings, Part I. editor / Vincente Julian ; David Camacho ; Hujun Yin ; Juan M. Alberola ; Vitor Beires Nogueira ; Paulo Novais ; Antonio Tallón-Ballesteros. Cham, Switzerland : Springer, 2024. pp. 155-166 (Lecture Notes in Computer Science).

@inproceedings{64ed31cf2c4a4db59668db1d0652faec,

title = "MetaLIRS: Meta-learning for Imputation and Regression Selection",

abstract = "Missing data is a prevalent problem in data science for many fields such as natural, social, and health sciences. Since most regression methods can not handle missing data directly, imputation methods are used in data pre-processing. Finding the best imputation method is non-trivial, however. Moreover, our results show that an independent choice for a best imputation method does not always result in the best predictive performance in the end; the combination matters. Furthermore, search-based approaches for finding a best-fitting imputer/regressor-pair can be computationally intensive. In this paper, we propose the MetaLIRS (Meta Learning Imputation and Regression Selection) frame-work for developing resource-friendly ML-based recommendation models for method selection. With MetaLIRS, we constructed a proof-of-concept recommendation model based on 12 meta-features that achieves an accuracy of 63\% for selecting the best-fitting imputer/regressor-pair. A data scientist can use this model for a quick resource-friendly recommendation on which imputation and regression method to use for their particular data set and task without the need for an expensive grid search among methods.",

keywords = "2024 OA procedure",

author = "\{Baysal Erez\}, I{\c s}ıl and Jan Flokstra and Mannes Poel and \{van Keulen\}, Maurice",

year = "2024",

month = nov,

day = "14",

doi = "10.1007/978-3-031-77731-8\_15",

language = "English",

isbn = "978-3-031-77730-1",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "155--166",

editor = "Vincente Julian and David Camacho and Hujun Yin and Alberola, \{Juan M.\} and \{Beires Nogueira\}, Vitor and Paulo Novais and Antonio Tall{\'o}n-Ballesteros",

booktitle = "Intelligent Data Engineering and Automated Learning - IDEAL 2024",

address = "Germany",

note = "25th International Conference on Intelligent Data Engineering and Automated Learning - IDEAL 2024, IDEAL 2024 ; Conference date: 20-11-2024 Through 22-11-2024",

}

Baysal Erez, I , Flokstra, J, Poel, M & van Keulen, M 2024, MetaLIRS: Meta-learning for Imputation and Regression Selection. in V Julian, D Camacho, H Yin, JM Alberola, V Beires Nogueira, P Novais & A Tallón-Ballesteros (eds), Intelligent Data Engineering and Automated Learning - IDEAL 2024: 25th International Conference, Valencia, Spain, November 20-22, 2024. Proceedings, Part I. Lecture Notes in Computer Science, vol. 15346, Springer, Cham, Switzerland, pp. 155-166, 25th International Conference on Intelligent Data Engineering and Automated Learning - IDEAL 2024, Valencia, Spain, 20/11/24. https://doi.org/10.1007/978-3-031-77731-8_15

MetaLIRS: Meta-learning for Imputation and Regression Selection. / Baysal Erez, Işıl ; Flokstra, Jan; Poel, Mannes et al.
Intelligent Data Engineering and Automated Learning - IDEAL 2024: 25th International Conference, Valencia, Spain, November 20-22, 2024. Proceedings, Part I. ed. / Vincente Julian; David Camacho; Hujun Yin; Juan M. Alberola; Vitor Beires Nogueira; Paulo Novais; Antonio Tallón-Ballesteros. Cham, Switzerland: Springer, 2024. p. 155-166 (Lecture Notes in Computer Science; Vol. 15346).

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

TY - GEN

T1 - MetaLIRS

T2 - 25th International Conference on Intelligent Data Engineering and Automated Learning - IDEAL 2024

AU - Baysal Erez, Işıl

AU - Flokstra, Jan

AU - Poel, Mannes

AU - van Keulen, Maurice

N1 - Conference code: 25

PY - 2024/11/14

Y1 - 2024/11/14

N2 - Missing data is a prevalent problem in data science for many fields such as natural, social, and health sciences. Since most regression methods can not handle missing data directly, imputation methods are used in data pre-processing. Finding the best imputation method is non-trivial, however. Moreover, our results show that an independent choice for a best imputation method does not always result in the best predictive performance in the end; the combination matters. Furthermore, search-based approaches for finding a best-fitting imputer/regressor-pair can be computationally intensive. In this paper, we propose the MetaLIRS (Meta Learning Imputation and Regression Selection) frame-work for developing resource-friendly ML-based recommendation models for method selection. With MetaLIRS, we constructed a proof-of-concept recommendation model based on 12 meta-features that achieves an accuracy of 63% for selecting the best-fitting imputer/regressor-pair. A data scientist can use this model for a quick resource-friendly recommendation on which imputation and regression method to use for their particular data set and task without the need for an expensive grid search among methods.

AB - Missing data is a prevalent problem in data science for many fields such as natural, social, and health sciences. Since most regression methods can not handle missing data directly, imputation methods are used in data pre-processing. Finding the best imputation method is non-trivial, however. Moreover, our results show that an independent choice for a best imputation method does not always result in the best predictive performance in the end; the combination matters. Furthermore, search-based approaches for finding a best-fitting imputer/regressor-pair can be computationally intensive. In this paper, we propose the MetaLIRS (Meta Learning Imputation and Regression Selection) frame-work for developing resource-friendly ML-based recommendation models for method selection. With MetaLIRS, we constructed a proof-of-concept recommendation model based on 12 meta-features that achieves an accuracy of 63% for selecting the best-fitting imputer/regressor-pair. A data scientist can use this model for a quick resource-friendly recommendation on which imputation and regression method to use for their particular data set and task without the need for an expensive grid search among methods.

KW - 2024 OA procedure

U2 - 10.1007/978-3-031-77731-8_15

DO - 10.1007/978-3-031-77731-8_15

M3 - Conference contribution

SN - 978-3-031-77730-1

T3 - Lecture Notes in Computer Science

SP - 155

EP - 166

BT - Intelligent Data Engineering and Automated Learning - IDEAL 2024

A2 - Julian, Vincente

A2 - Camacho, David

A2 - Yin, Hujun

A2 - Alberola, Juan M.

A2 - Beires Nogueira, Vitor

A2 - Novais, Paulo

A2 - Tallón-Ballesteros, Antonio

PB - Springer

CY - Cham, Switzerland

Y2 - 20 November 2024 through 22 November 2024

ER -

Baysal Erez I , Flokstra J, Poel M, van Keulen M. MetaLIRS: Meta-learning for Imputation and Regression Selection. In Julian V, Camacho D, Yin H, Alberola JM, Beires Nogueira V, Novais P, Tallón-Ballesteros A, editors, Intelligent Data Engineering and Automated Learning - IDEAL 2024: 25th International Conference, Valencia, Spain, November 20-22, 2024. Proceedings, Part I. Cham, Switzerland: Springer. 2024. p. 155-166. (Lecture Notes in Computer Science). doi: 10.1007/978-3-031-77731-8_15

MetaLIRS: Meta-learning for Imputation and Regression Selection

Abstract

Publication series

Conference

Keywords

Access to Document

Fingerprint

Cite this