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

Consistency and reproducibility of attribute extraction by different machine learning systems.

Zoran Lukacic1, Josipa Kern

  • 1Heart Rehabilitation Centre, Kidriceva 6, 9252 Radenci, Slovenia. zoran.lukacic@siol.net

Studies in Health Technology and Informatics
|October 6, 2004
PubMed
Summary

Two machine learning (ML) systems, See5 and FMLS, were compared for attribute extraction from medical data. Both systems demonstrated intra-testing consistency but lacked inter-testing consistency in extracted attributes.

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

Is artificial intelligence really a new topic in medical education?

Croatian medical journal·2021
Same author

Primary School Teachers' Knowledge on Tooth Avulsion.

Acta stomatologica Croatica·2021
Same author

Smoking Behaviors in Kosova: Results of Steps Survey.

Zdravstveno varstvo·2017
Same author

Missing risks in opportunistic screening for type 2 diabetes - CroDiabGP study.

Medicinski glasnik : official publication of the Medical Association of Zenica-Doboj Canton, Bosnia and Herzegovina·2016
Same author

Development of the quality assessment model of EHR software in family medicine practices: research based on user satisfaction.

Journal of innovation in health informatics·2015
Same author

Recipe from the past for the future: public health intervention represent a process for century.

Materia socio-medica·2015

Area of Science:

  • Computer Science
  • Medical Informatics
  • Data Science

Background:

  • Machine learning (ML) systems are crucial for intelligent data analysis and attribute extraction for outcome prediction.
  • Evaluating the consistency and reproducibility of these extracted attributes is essential for reliable ML applications, particularly in sensitive fields like medicine.

Purpose of the Study:

  • To compare the performance of two ML systems, See5 and FMLS, in extracting relevant attributes from a real medical dataset.
  • To propose a method for testing the intra- and inter-testing consistency and reproducibility of attributes extracted by these ML systems.

Main Methods:

  • Utilized two ML systems, See5 and FMLS, to extract relevant attributes from a real-world medical dataset.
  • Performed intra- and inter-testing consistency evaluations on the extracted attributes.

Related Experiment Videos

  • Compared the accuracy, sensitivity, and specificity of both systems under different attribute extraction conditions.

    • Main Results:

    • Accuracy and sensitivity were comparable when FMLS used attributes extracted by See5; slightly lower when FMLS used its own attributes.
    • Specificity was similar when FMLS was guided by See5's attributes but significantly higher when FMLS used its own.
    • Both ML systems exhibited strong intra-testing consistency but failed to demonstrate inter-testing consistency.

    Conclusions:

    • See5 and FMLS show comparable performance in attribute extraction for medical data, with variations in specificity depending on the extraction method.
    • The proposed testing method highlights that while ML systems can be consistent within a single test (intra-testing), they may not yield reproducible results across different tests (inter-testing).
    • Further research is needed to address the inter-testing inconsistency observed in attribute extraction for robust medical data analysis.