Conformal Rule-Based Multi-label Classification

Eyke Hüllermeier; Johannes Fürnkranz; Eneldo Loza Mencía

Conformal Rule-Based Multi-label Classification

Eyke Hüllermeier, Johannes Fürnkranz, Eneldo Loza Mencía

Institute for Application-oriented Knowledge Processing

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

Abstract

We advocate the use of conformal prediction (CP) to enhance rule-based multi-label classification (MLC). In particular, we highlight the mutual benefit of CP and rule learning: Rules have the ability to provide natural (non-)conformity scores, which are required by CP, while CP suggests a way to calibrate the assessment of candidate rules, thereby supporting better predictions and more elaborate decision making. We illustrate the potential usefulness of calibrated conformity scores in a case study on lazy multi-label rule learning.

Original language	English
Title of host publication	Proceedings of the 43d German Conference on Artificial Intelligence (KI-20)
Editors	Ute Schmid and Diedrich Wolter and Franziska Klügl
Place of Publication	Bamberg, Germany
Publisher	Springer
Pages	290-296
Number of pages	7
Volume	12325
ISBN (Print)	978-3-030-58284-5
Publication status	Published - 2020

Publication series

Name	Lecture Notes in Computer Science (LNCS)

Fields of science

102001 Artificial intelligence
102019 Machine learning
102028 Knowledge engineering
102033 Data mining

JKU Focus areas

Digital Transformation

Cite this

@inproceedings{bd6252d890b74ca08367f450bc79488b,

title = "Conformal Rule-Based Multi-label Classification",

abstract = "We advocate the use of conformal prediction (CP) to enhance rule-based multi-label classification (MLC). In particular, we highlight the mutual benefit of CP and rule learning: Rules have the ability to provide natural (non-)conformity scores, which are required by CP, while CP suggests a way to calibrate the assessment of candidate rules, thereby supporting better predictions and more elaborate decision making. We illustrate the potential usefulness of calibrated conformity scores in a case study on lazy multi-label rule learning.",

author = "Eyke H{\"u}llermeier and Johannes F{\"u}rnkranz and {Loza Menc{\'i}a}, Eneldo",

year = "2020",

language = "English",

isbn = "978-3-030-58284-5",

volume = "12325",

series = "Lecture Notes in Computer Science (LNCS)",

publisher = "Springer",

pages = "290--296",

editor = "{Ute Schmid and Diedrich Wolter and Franziska Kl{\"u}gl}",

booktitle = "Proceedings of the 43d German Conference on Artificial Intelligence (KI-20)",

}

Conformal Rule-Based Multi-label Classification. / Hüllermeier, Eyke; Fürnkranz, Johannes; Loza Mencía, Eneldo.
Proceedings of the 43d German Conference on Artificial Intelligence (KI-20). ed. / Ute Schmid and Diedrich Wolter and Franziska Klügl. Vol. 12325 Bamberg, Germany: Springer, 2020. p. 290-296 (Lecture Notes in Computer Science (LNCS)).

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

TY - GEN

T1 - Conformal Rule-Based Multi-label Classification

AU - Hüllermeier, Eyke

AU - Fürnkranz, Johannes

AU - Loza Mencía, Eneldo

PY - 2020

Y1 - 2020

N2 - We advocate the use of conformal prediction (CP) to enhance rule-based multi-label classification (MLC). In particular, we highlight the mutual benefit of CP and rule learning: Rules have the ability to provide natural (non-)conformity scores, which are required by CP, while CP suggests a way to calibrate the assessment of candidate rules, thereby supporting better predictions and more elaborate decision making. We illustrate the potential usefulness of calibrated conformity scores in a case study on lazy multi-label rule learning.

AB - We advocate the use of conformal prediction (CP) to enhance rule-based multi-label classification (MLC). In particular, we highlight the mutual benefit of CP and rule learning: Rules have the ability to provide natural (non-)conformity scores, which are required by CP, while CP suggests a way to calibrate the assessment of candidate rules, thereby supporting better predictions and more elaborate decision making. We illustrate the potential usefulness of calibrated conformity scores in a case study on lazy multi-label rule learning.

M3 - Conference proceedings

SN - 978-3-030-58284-5

VL - 12325

T3 - Lecture Notes in Computer Science (LNCS)

SP - 290

EP - 296

BT - Proceedings of the 43d German Conference on Artificial Intelligence (KI-20)

A2 - Ute Schmid and Diedrich Wolter and Franziska Klügl, null

PB - Springer

CY - Bamberg, Germany

ER -