Null QQ plots: A simple graphical alternative to significance testing for the comparison of classifiers

Daniel Berrar

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The evaluation of machine learning algorithms is commonly based on statistical significance tests. However, the suitability of such tests is often questionable. We propose null QQ plots as a simple yet powerful graphical alternative to significance testing. Using ten benchmark data sets, we demonstrate that these plots concisely summarize the essential results from a comparative classification study, while they are easy to produce and interpret.

Original languageEnglish
Title of host publicationProceedings - International Conference on Pattern Recognition
Pages1852-1855
Number of pages4
Publication statusPublished - 2012
Externally publishedYes
Event21st International Conference on Pattern Recognition, ICPR 2012 - Tsukuba, Japan
Duration: 2012 Nov 112012 Nov 15

Other

Other21st International Conference on Pattern Recognition, ICPR 2012
CountryJapan
CityTsukuba
Period12/11/1112/11/15

Fingerprint

Statistical tests
Learning algorithms
Learning systems
Classifiers
Testing

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Cite this

Berrar, D. (2012). Null QQ plots: A simple graphical alternative to significance testing for the comparison of classifiers. In Proceedings - International Conference on Pattern Recognition (pp. 1852-1855). [6460514]

Null QQ plots : A simple graphical alternative to significance testing for the comparison of classifiers. / Berrar, Daniel.

Proceedings - International Conference on Pattern Recognition. 2012. p. 1852-1855 6460514.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Berrar, D 2012, Null QQ plots: A simple graphical alternative to significance testing for the comparison of classifiers. in Proceedings - International Conference on Pattern Recognition., 6460514, pp. 1852-1855, 21st International Conference on Pattern Recognition, ICPR 2012, Tsukuba, Japan, 12/11/11.
Berrar D. Null QQ plots: A simple graphical alternative to significance testing for the comparison of classifiers. In Proceedings - International Conference on Pattern Recognition. 2012. p. 1852-1855. 6460514
Berrar, Daniel. / Null QQ plots : A simple graphical alternative to significance testing for the comparison of classifiers. Proceedings - International Conference on Pattern Recognition. 2012. pp. 1852-1855
@inproceedings{7b2dc2fca71d4276a6cfb67de86d7012,
title = "Null QQ plots: A simple graphical alternative to significance testing for the comparison of classifiers",
abstract = "The evaluation of machine learning algorithms is commonly based on statistical significance tests. However, the suitability of such tests is often questionable. We propose null QQ plots as a simple yet powerful graphical alternative to significance testing. Using ten benchmark data sets, we demonstrate that these plots concisely summarize the essential results from a comparative classification study, while they are easy to produce and interpret.",
author = "Daniel Berrar",
year = "2012",
language = "English",
isbn = "9784990644109",
pages = "1852--1855",
booktitle = "Proceedings - International Conference on Pattern Recognition",

}

TY - GEN

T1 - Null QQ plots

T2 - A simple graphical alternative to significance testing for the comparison of classifiers

AU - Berrar, Daniel

PY - 2012

Y1 - 2012

N2 - The evaluation of machine learning algorithms is commonly based on statistical significance tests. However, the suitability of such tests is often questionable. We propose null QQ plots as a simple yet powerful graphical alternative to significance testing. Using ten benchmark data sets, we demonstrate that these plots concisely summarize the essential results from a comparative classification study, while they are easy to produce and interpret.

AB - The evaluation of machine learning algorithms is commonly based on statistical significance tests. However, the suitability of such tests is often questionable. We propose null QQ plots as a simple yet powerful graphical alternative to significance testing. Using ten benchmark data sets, we demonstrate that these plots concisely summarize the essential results from a comparative classification study, while they are easy to produce and interpret.

UR - http://www.scopus.com/inward/record.url?scp=84874581269&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84874581269&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84874581269

SN - 9784990644109

SP - 1852

EP - 1855

BT - Proceedings - International Conference on Pattern Recognition

ER -