Fuzzy-rough nearest neighbour classification and prediction

  • Authors:
  • Richard Jensen;Chris Cornelis

  • Affiliations:
  • Department of Computer Science, Aberystwyth University, Ceredigion, SY23 3DB, Wales, UK;Department of Applied Mathematics and Computer Science, Ghent University, Gent, Belgium

  • Venue:
  • Theoretical Computer Science
  • Year:
  • 2011

Quantified Score

Hi-index 5.23

Visualization

Abstract

Nearest neighbour (NN) approaches are inspired by the way humans make decisions, comparing a test object to previously encountered samples. In this paper, we propose an NN algorithm that uses the lower and upper approximations from fuzzy-rough set theory in order to classify test objects, or predict their decision value. It is shown experimentally that our method outperforms other NN approaches (classical, fuzzy and fuzzy-rough ones) and that it is competitive with leading classification and prediction methods. Moreover, we show that the robustness of our methods against noise can be enhanced effectively by invoking the approximations of the Vaguely Quantified Rough Set (VQRS) model, which emulates the linguistic quantifiers ''some'' and ''most'' from natural language.