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

Integrating regression formulas and kernel functions into locally adaptive knowledge-based neural networks: a case

Qun Song1, Nikola Kasabov, Tianmin Ma

  • 1Knowledge Engineering & Discovery Research Institute, Auckland University of Technology, Private Bag 92006, Auckland 1020, New Zealand.

Artificial Intelligence in Medicine
|October 11, 2005
PubMed
Summary

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

Local global feature enhanced transformer with attention pruning for precise medical image segmentation.

iScience·2026
Same author

Tunable Interference Colors in Nanofibril-Crystal Composite Films via Integrated Salt-Assisted Assembly.

Research (Washington, D.C.)·2026
Same author

MsMemoryGAN: A Multiscale Memory GAN for Palm-Vein Adversarial Purification.

IEEE transactions on cybernetics·2026
Same author

Comparative performance analysis of quantum feature maps for quantum kernel-based machine learning.

Scientific reports·2026
Same author

An integrated AI-enabled system using One Class Twin Cross Learning for early gastric cancer detection.

Frontiers in oncology·2026
Same author

The protection of mesenchymal stem cells in metabolic reprogramming and endothelial-mesenchymal transition in diabetic aortas.

Stem cells translational medicine·2026

A new knowledge-based neural network model (KBNN) improves renal function evaluation by integrating multiple regression formulas. This AI model achieves higher accuracy than existing methods, offering localized predictions and explanatory rules for better patient care.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Medical Informatics

Background:

  • Existing regression formulas for medical outcome prediction are often limited to specific data subspaces.
  • A unified approach is needed to leverage multiple formulas for improved accuracy and broader applicability.
  • Renal function evaluation currently relies on various predictive formulas, each with limitations.

Purpose of the Study:

  • To develop a generic, incremental learning model that incorporates multiple regression formulas for enhanced prediction.
  • To define local areas within the problem space and identify the best-performing formula for each.
  • To create a specific model for renal function evaluation using nine established formulas.

Main Methods:

  • Developed a knowledge-based neural network (KBNN) model using a connectionist neuro-fuzzy approach.

Related Experiment Videos

  • Incorporated non-linear regression functions and Gaussian kernel functions within hidden neural nodes for localized adaptation.
  • Utilized incremental learning to adapt functions based on data within defined subspaces, tested on a renal function dataset (GFR).

    • Main Results:

    • The KBNN model demonstrated over 10% higher accuracy in Glomerular Filtration Rate (GFR) prediction compared to individual formulas or standard neural networks.
    • Derived locally adapted regression formulas optimized for specific data clusters.
    • Generated useful explanatory rules alongside the localized predictions.

    Conclusions:

    • The proposed KBNN model offers superior accuracy for renal function evaluation compared to existing methods.
    • The model successfully extracts modified formulas that perform optimally in local data areas.
    • This approach provides a more robust and interpretable framework for medical outcome prediction.