Learning Optimal Parameters in Decision-Theoretic Rough Sets

  • Authors:
  • Joseph P. Herbert;Jingtao Yao

  • Affiliations:
  • Department of Computer Science, University of Regina, Regina, Canada S4S 0A2;Department of Computer Science, University of Regina, Regina, Canada S4S 0A2

  • Venue:
  • RSKT '09 Proceedings of the 4th International Conference on Rough Sets and Knowledge Technology
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

A game-theoretic approach for learning optimal parameter values for probabilistic rough set regions is presented. The parameters can be used to define approximation regions in a probabilistic decision space. New values for loss functions are learned from a sequence of risk modifications derived from game-theoretic analysis of the relationship between two classification measures. Using game theory to maximize these measures results in a learning method to reformulate the loss functions. The decision-theoretic rough set model acquires initial values for these parameters through a combination of loss functions provided by the user. The new game-theoretic learning method modifies these loss functions according to an acceptable threshold.