A statistical technique for comparing the accuracies of several classifiers

Research output: Contribution to journalArticle

23 Scopus citations


The problem of comparing the accuracies of several classifiers that have been applied to the same set of observations is considered. A strategy based on the application of repeated measures analysis procedures to dichotomous data is proposed for making the comparison. The strategy is illustrated using a set of data taken from the literature.

Original languageEnglish (US)
Pages (from-to)5-9
Number of pages5
JournalPattern Recognition Letters
Issue number1
Publication statusPublished - Jul 1988
Externally publishedYes



  • Cochran Q test
  • Related proportions
  • adjusted F tests
  • pairwise comparisons
  • repeated measures
  • training set
  • validation set

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Cite this