Performance advantage of combined classifiers in multi-category cases: An analysis

Xubo Song, Misha Pavel

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

One problem in the field of machine learning is that the performance on the training and validation sets lack robustness when applied in real-life situations. Recent advances in ensemble methods have demonstrated that robust behavior can be improved by combining a large number of weak classifiers. The key insight of this paper is that the performance enhancement due to combining multiple classifiers is considerably greater in multi-category situations than in binary classifications, as long as their errors arc conditionally independent. This paper provides some experimental and theoretical analysis of the performance using majority vote, paying special attention to the effect of several parameters that include the number of combined classifiers, weakness of the combined classifiers and the number of classes. These insights can provide guidance for the analysis and design of multi-classifier systems.

Original languageEnglish (US)
Pages (from-to)750-757
Number of pages8
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume3316
StatePublished - Dec 1 2004

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Performance advantage of combined classifiers in multi-category cases: An analysis'. Together they form a unique fingerprint.

Cite this