Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • harvard-cite-them-right
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Evolving Accurate and Comprehensible Classification Rules
University of Borås, School of Business and IT. (CSL@BS)
University of Borås, School of Business and IT. (CSL@BS)
University of Borås, School of Business and IT. (CSL@BS)
2011 (English)Conference paper, Published paper (Refereed)
Abstract [en]

In this paper, Genetic Programming is used to evolve ordered rule sets (also called decision lists) for a number of benchmark classification problems, with evaluation of both predictive performance and comprehensibility. The main purpose is to compare this approach to the standard decision list algorithm JRip and also to evaluate the use of different length penalties and fitness functions for evolving this type of model. The results, using 25 data sets from the UCI repository, show that genetic decision lists with accuracy-based fitness functions outperform JRip regarding accuracy. Indeed, the best setup was significantly better than JRip. JRip, however, held a slight advantage over these models when evaluating AUC. Furthermore, all genetic decision list setups produced models that were more compact than JRip models, and thus more readily comprehensible. The effect of using different fitness functions was very clear; in essence, models performed best on the evaluation criterion that was used in the fitness function, with a worsening of the performance for other criteria. Brier score fitness provided a middle ground, with acceptable performance on both accuracy and AUC. The main conclusion is that genetic programming solves the task of evolving decision lists very well, but that different length penalties and fitness functions have immediate effects on the results. Thus, these parameters can be used to control the trade-off between different aspects of predictive performance and comprehensibility.

Place, publisher, year, edition, pages
IEEE , 2011.
Keywords [en]
genetic programming, decision lists, Machine Learning
Keywords [sv]
data mining, Data Mining
National Category
Computer and Information Sciences Information Systems
Research subject
Bussiness and IT
Identifiers
URN: urn:nbn:se:hb:diva-6698DOI: 10.1109/CEC.2011.5949784Local ID: 2320/10007ISBN: 978-1-4244-7834-7 (print)OAI: oai:DiVA.org:hb-6698DiVA, id: diva2:887399
Conference
IEEE Congress on Evolutionary Computation (CEC), 5-8 juni, New orleans, LA, USA, 2011
Available from: 2015-12-22 Created: 2015-12-22 Last updated: 2018-01-10

Open Access in DiVA

fulltext(118 kB)706 downloads
File information
File name FULLTEXT01.pdfFile size 118 kBChecksum SHA-512
a57a7e775aa6a3862211d2c529416ba9c969f636f17da9c3472ac4a90d459578f95f15c4f073b673a93bee4b94cb90580b8bb739d1461b2e6c5397b4bf4abf2e
Type fulltextMimetype application/pdf

Other links

Publisher's full text

Authority records

Sönströd, CeciliaJohansson, UlfKönig, Rikard

Search in DiVA

By author/editor
Sönströd, CeciliaJohansson, UlfKönig, Rikard
By organisation
School of Business and IT
Computer and Information SciencesInformation Systems

Search outside of DiVA

GoogleGoogle Scholar
Total: 706 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 215 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • harvard-cite-them-right
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf