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

Nearest neighbour classification with heterogeneous proximity functions.

J Laurikkala1, M Juhola

  • 1Department of Computer Science, 33014 University of Tampere, Finland. jpl@cs.uta.fi

Studies in Health Technology and Informatics
|February 24, 2001
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

Application of a modified two-point backward difference to sequential event detection in surface electromyography.

Journal of medical engineering & technology·2009
Same author

Modelling stabilograms with hidden Markov models.

Journal of medical engineering & technology·2008
Same author

Differential cancer predisposition in Lynch syndrome: insights from molecular analysis of brain and urinary tract tumors.

Carcinogenesis·2008
Same author

Is gastric cancer part of the tumour spectrum of hereditary non-polyposis colorectal cancer? A molecular genetic study.

Gut·2007
Same author

CD44 expression and its relationship with MMP-9, clinicopathological factors and survival in oral squamous cell carcinoma.

Oral oncology·2006
Same author

Is it possible to reduce endoscopy workload using age, alarm symptoms and H. pylori as predictors of peptic ulcer and oesophagogastric cancers?

Digestive and liver disease : official journal of the Italian Society of Gastroenterology and the Italian Association for the Study of the Liver·2005

Heterogeneous proximity functions were compared for mixed-type data classification. The Heterogeneous Value Difference Metric (HVDM) demonstrated superior performance, effectively handling nominal attributes for better nearest neighbour classification accuracy.

Area of Science:

  • Data Science
  • Machine Learning
  • Pattern Recognition

Background:

  • Proximity functions are crucial for measuring similarity or distance between data points.
  • Traditional functions may struggle with datasets containing attributes of mixed types (numerical and categorical).
  • Heterogeneous proximity functions offer a potential solution by adapting processing based on attribute scale.

Purpose of the Study:

  • To evaluate the effectiveness of heterogeneous proximity functions against standard Minkowskian distances for mixed-type data.
  • To identify which proximity function yields the highest nearest neighbour classification accuracy on datasets with mixed attributes.

Main Methods:

  • Comparison of two Minkowskian distance functions (City-block, Euclidean) with three heterogeneous proximity functions (Gower's similarity, Heterogeneous Euclidean-Overlap Metric, Heterogeneous Value Difference Metric - HVDM).

Related Experiment Videos

  • Evaluation conducted on 21 datasets characterized by mixed attribute types.
  • Nearest neighbour classification accuracy was the primary performance metric.

    • Main Results:

    • Significant differences in classification accuracy were observed across 11 of the 21 datasets.
    • City-block and Euclidean distances outperformed Gower's similarity and Heterogeneous Euclidean-Overlap Metric.
    • The Heterogeneous Value Difference Metric (HVDM) significantly outperformed all other tested functions.

    Conclusions:

    • HVDM is the most effective proximity function for classifying datasets with mixed attribute types.
    • HVDM's superior performance stems from its careful handling of nominal attributes.
    • The study highlights the importance of specialized functions for heterogeneous data in machine learning tasks.