Relationship between diversity and correlation in Multi-Classifier Systems

Kuo Wei Hsu, Jaideep Srivastava

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Scopus citations

Abstract

Diversity plays an important role in the design of Multi-Classifier Systems, but its relationship to classification accuracy is still unclear from a theoretical perspective. As a step towards the solution of this probelm, we take a different route and explore the relationship between diversity and correlation. In this paper we provide a theoretical analysis and present a nonlinear function that relates diversity to correlation, which hence can be further related to accuracy. This paper contributes to connecting existing research in diversity and correlation, and also providing a proxy to the relationship between diversity and accuracy. Our experimental results reveal deeper insights into the role of diversity in Multi-Classifier Systems.

Original languageEnglish (US)
Title of host publicationAdvances in Knowledge Discovery and Data Mining - 14th Pacific-Asia Conference, PAKDD 2010, Proceedings
Pages500-506
Number of pages7
EditionPART 2
DOIs
StatePublished - Dec 1 2010
Event14th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2010 - Hyderabad, India
Duration: Jun 21 2010Jun 24 2010

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume6119 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other14th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2010
CountryIndia
CityHyderabad
Period6/21/106/24/10

Keywords

  • Correlation
  • Diversity
  • Multi-Classifier System (MCS)

Fingerprint

Dive into the research topics of 'Relationship between diversity and correlation in Multi-Classifier Systems'. Together they form a unique fingerprint.

Cite this