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

Classification-algorithm evaluation: five performance measures based on confusion matrices

A D Forbes1

  • 1Medical Department, Hewlett-Packard Laboratories, Palo Alto, CA 94303-0867, USA.

Journal of Clinical Monitoring
|May 1, 1995
PubMed
Summary
This summary is machine-generated.

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

Algorithms and the Journal of Clinical Monitoring--II.

Journal of clinical monitoring·1994
Same author

Inhibition of neutrophil adhesion does not prevent ischemic spinal cord injury.

The Annals of thoracic surgery·1994
Same author

Mechanical circulatory support during repair of thoracic aortic injuries improves morbidity and prevents spinal cord injury.

Archives of surgery (Chicago, Ill. : 1960)·1994
Same author

Schwannoma of the left atrium: diagnostic evaluation and surgical resection.

The Annals of thoracic surgery·1994
Same author

Acquired immunodeficiency syndrome-related lymphoma of the lung presenting as empyema thoracis.

The Annals of thoracic surgery·1994
Same author

Echocardiographic volume flow and stenosis severity measures with changing flow rate in aortic stenosis.

The American journal of physiology·1993
Same journal

Logistics of conducting a large number of individual sessions with a full-scale patient simulator at a scientific meeting.

Journal of clinical monitoring·1998
Same journal

Electrocardiogram trace paradoxically filtered by activation of an electrosurgical unit.

Journal of clinical monitoring·1998
Same journal

A robust sensor fusion method for heart rate estimation.

Journal of clinical monitoring·1998
Same journal

Robust sensor fusion improves heart rate estimation: clinical evaluation.

Journal of clinical monitoring·1998
Same journal

Hemocue, an accurate bedside method of hemoglobin measurement?

Journal of clinical monitoring·1998
Same journal

A new technique to measure and track blood resistivity in intracardiac impedance volumetry.

Journal of clinical monitoring·1998
See all related articles

This study introduces a new method for comparing classification algorithm performance using normalized mutual information (NMI). The modified NMI is preferred for its adherence to constraints and conservative performance evaluation.

Area of Science:

  • Computer Science
  • Machine Learning
  • Data Science

Background:

  • Classification algorithms are crucial in data analysis.
  • Evaluating algorithm performance requires robust metrics.
  • Traditional performance measures have limitations.

Purpose of the Study:

  • To introduce, explain, and extend methods for comparing classification algorithm performance.
  • To address limitations in traditional performance evaluation metrics.
  • To propose a novel, reliable measure for algorithm comparison.

Main Methods:

  • Defined two contexts: objects-by-inspection and objects-by-segmentation.
  • Utilized confusion matrices for performance assessment.
  • Critiqued four traditional performance measures against defined requirements and constraints.

Related Experiment Videos

  • Introduced the normalized mutual information (NMI) as a novel performance measure.

    • Main Results:

    • Traditional measures met requirements but failed to obey constraints.
    • The normalized mutual information (NMI) was introduced.
    • Methods for comparing algorithm performance using confusion matrices based on NMI were devised.
    • Five performance measures yielded similar inferences for QRS-detection algorithms.

    Conclusions:

    • The modified NMI is the preferred measure for comparing classification algorithm performance.
    • The modified NMI adheres to all defined constraints.
    • It offers the most conservative evaluation of algorithm performance.