Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Comparison of the ID3 algorithm versus discriminant analysis for performing feature selection.

E L Kinney1, D D Murphy

  • 1Cardiology Division, University of Miami School of Medicine, Florida.

Computers and Biomedical Research, an International Journal
|October 1, 1987
PubMed
Summary

This study compared ID3 and discriminant analysis for medical data classification. Both algorithms performed poorly, with ID3 showing optimistic accuracy estimates and issues with larger datasets.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Extinction, reduction, stability and increase: The responses of checkerspot butterfly (Euphydryas) populations to the California drought.

Oecologia·2017
Same author

Population structure of a hilltopping butterfly.

Oecologia·2017
Same author

A reexamination of hilltopping in Euphydryas editha.

Oecologia·2011
Same author

Neuroprotective agents for clinical trials in Parkinson's disease: a systematic assessment.

Neurology·2003
Same author

Transfected synphilin-1 forms cytoplasmic inclusions in HEK293 cells.

Brain research. Molecular brain research·2001
Same author

Feasibility and acceptability of the COOP/WONCA charts for identification of functional limitations in rural patients of the People's Republic of China.

International journal of rehabilitation research. Internationale Zeitschrift fur Rehabilitationsforschung. Revue internationale de recherches de readaptation·2001

Area of Science:

  • Medical Informatics
  • Machine Learning in Healthcare

Background:

  • Initial attempts using the ID3 algorithm on a small medical dataset yielded disappointing results.
  • The dataset, related to cardiology auscultation accuracy, was initially deemed suitable for ID3 induction.

Purpose of the Study:

  • To compare the classification performance of the ID3 algorithm against discriminant analysis.
  • To evaluate the reliability of ID3's probability estimates for future case classification.
  • To assess algorithm performance with increasing sample sizes.

Main Methods:

  • Comparative analysis of ID3 and discriminant analysis.
  • Performance evaluation using percentage of correct classification on an independent dataset (n=67).
  • Assessment of algorithm behavior with increased sample size.

Related Experiment Videos

Main Results:

  • Both ID3 (60% correct) and discriminant analysis (66% correct) exhibited poor performance on the independent dataset.
  • ID3's probability statistics were overly optimistic compared to actual results.
  • Increasing sample size led to a complex, less generalizable decision tree for ID3, while discriminant analysis maintained accuracy.

Conclusions:

  • ID3 algorithm demonstrates limitations in medical data classification tasks, particularly with larger datasets.
  • Discriminant analysis, while not optimal, showed more stable performance characteristics than ID3.
  • More sophisticated algorithms are needed for medical data analysis, even if computationally less efficient than ID3.