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

Efficient kNN Classification With Different Numbers of Nearest Neighbors.

Shichao Zhang, Xuelong Li, Ming Zong

    IEEE Transactions on Neural Networks and Learning Systems
    |April 20, 2017
    PubMed
    Summary

    The novel kTree and k*Tree methods enhance nearest neighbor (kNN) classification by learning optimal k values for each sample, improving accuracy and efficiency over traditional fixed-value or cross-validation approaches.

    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

    Analysis of ultrasonic characteristics in 12 cases of ovarian Sertoli-Leydig cell tumour.

    Frontiers in oncology·2026
    Same author

    Simulation and Experimental Study on Parameter Optimization for the Glass Molding Process of Automotive Panoramic Roofs.

    Materials (Basel, Switzerland)·2026
    Same author

    Pre-implantation embryo metabolism identified by PEMA reveals endogenous lactate insufficiency contributes to pre-implantation development arrest.

    Fundamental research·2026
    Same author

    Dual-Modal Serum and Urine SERS Metabolic Fingerprint for the Diagnosis and Activity Assessment of Childhood Lupus Nephritis.

    Analytical chemistry·2026
    Same author

    Microencapsulated Glass Ionomer Cement-Driven Dental Self-Healing Resin Composites With Enhanced Mechanical Strength, Biocompatibility, and Reparative Dentin Formation.

    Advanced healthcare materials·2026
    Same author

    SIRL-1 deficiency reveals pro-inflammatory IL-8 axis in inflammatory bowel disease: a novel diagnostic ratio.

    Frontiers in immunology·2026

    Area of Science:

    • Machine Learning
    • Data Mining
    • Statistical Classification

    Background:

    • Traditional nearest neighbor (kNN) methods use a fixed k value, which is suboptimal for diverse datasets.
    • Existing methods for optimizing k are often time-consuming, hindering practical application.

    Purpose of the Study:

    • To propose an efficient kNN classification method that learns optimal k values for individual samples.
    • To introduce the kTree and k*Tree algorithms for improved classification accuracy and reduced computational cost.

    Main Methods:

    • The kTree method involves a training stage to learn optimal k values using a sparse reconstruction model and constructs a decision tree.
    • The k*Tree method enhances kTree by storing additional training sample information in leaf nodes for faster testing.

    Related Experiment Videos

  • Both methods utilize the learned optimal k values during the test stage for kNN classification.

    • Main Results:

    • kTree achieves higher classification accuracy than traditional kNN methods with similar running costs.
    • k*Tree offers reduced running costs compared to other adaptive kNN methods while maintaining similar accuracy.
    • Experimental results on 20 datasets demonstrate the superior efficiency of kTree and k*Tree.

    Conclusions:

    • The proposed kTree and k*Tree methods provide efficient and accurate solutions for kNN classification by adaptively determining the optimal k value.
    • These methods offer a significant improvement over existing approaches, particularly in terms of computational efficiency during the test phase.
    • The findings suggest broader applicability of these adaptive kNN techniques in data mining and statistical classification tasks.