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

Profiling local optima in K-means clustering: developing a diagnostic technique.

Douglas Steinley1

  • 1Department of Psychological Sciences, University of Missouri-Columbia, Columbia, MO 65203, USA. steinleyd@missouri.edu

Psychological Methods
|June 21, 2006
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

Right-sizing growth mixture models as multi-group growth and confirmatory factor models.

Behavior research methods·2025
Same author

Cognitive behavioral therapy for insomnia among heavy-drinking veterans: a randomized pilot trial.

Sleep advances : a journal of the Sleep Research Society·2025
Same author

"General Addiction Liability" Revisited.

Clinical psychological science : a journal of the Association for Psychological Science·2025
Same author

Integer programming in psychology: A review and directions for future research.

The British journal of mathematical and statistical psychology·2025
Same author

Daily associations between sleep and alcohol use among veterans: Acute and cumulative effects.

Addiction (Abingdon, England)·2025
Same author

Clustering methods: To optimize or to not optimize?

Psychological methods·2024
Same journal

Bayesian evaluation for latent variable models: A tutorial on computing information criteria and bayes factors with the r package bleval.

Psychological methods·2026
Same journal

A stochastic block prior for clustering in graphical models.

Psychological methods·2026
Same journal

Three-level vector autoregressive models.

Psychological methods·2026
Same journal

Scaling cognitive modeling to big data: A deep learning approach to studying individual differences in evidence accumulation model parameters.

Psychological methods·2026
Same journal

Best practices in multilevel modeling for within-cluster group comparisons: An evaluation of coding strategies reflecting group composition and heterogeneity.

Psychological methods·2026
Same journal

A unified framework for psychometrics in experimental psychology: The standardized generalized hierarchical factor model.

Psychological methods·2026
See all related articles

This study evaluates K-means clustering performance across various data scenarios, including cluster overlap and distributions. A new diagnostic technique helps determine when K-means results are reliable for data analysis.

Area of Science:

  • Data Science
  • Machine Learning
  • Statistical Analysis

Background:

  • K-means clustering is a widely used algorithm for partitioning data.
  • Its performance can be sensitive to data characteristics like cluster overlap and distribution.
  • Assessing the reliability of K-means results is crucial for data-driven decision-making.

Purpose of the Study:

  • To systematically investigate the performance of K-means clustering under diverse conditions.
  • To develop a diagnostic method for evaluating the trustworthiness of K-means partitioning.
  • To provide guidance for key data analysis decisions based on K-means output.

Main Methods:

  • Utilized the cluster generation procedure by Steinley and Henson (2005).
  • Evaluated K-means performance across scenarios including cluster overlap probability and type, sample size, dimensions, cluster distributions, and data structures.

Related Experiment Videos

  • Employed the Hubert-Arabie adjusted Rand index for result evaluation.
  • Implemented thousands of K-means algorithm runs to assess local optima prevalence.

    • Main Results:

    • Identified key factors influencing K-means clustering performance, such as cluster overlap and data dimensionality.
    • Observed specific patterns in K-means behavior under different multivariate distributions.
    • Demonstrated the impact of sample size and number of clusters on partitioning accuracy.

    Conclusions:

    • K-means clustering performance varies significantly with data characteristics.
    • A novel diagnostic technique is proposed to assess the reliability of K-means partitioning.
    • This method combines data features with algorithm behavior to guide data analysis choices.