Inducing Better Rule Sets by Adding Missing Attribute Values

  • Authors:
  • Jerzy W. Grzymala-Busse;Witold J. Grzymala-Busse

  • Affiliations:
  • Department of Electrical Engineering and Computer Science, University of Kansas, Lawrence, USA KS 66045 and Institute of Computer Science, Polish Academy of Sciences, Warsaw, Poland 01---237;Touchnet Information Systems, Inc., Lenexa, USA KS 66219

  • Venue:
  • RSCTC '08 Proceedings of the 6th International Conference on Rough Sets and Current Trends in Computing
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Our main objective was to verify the following hypothesis: for some complete (i.e., without missing attribute vales) data sets it is possible to induce better rule sets (in terms of an error rate) by increasing incompleteness (i.e., removing some existing attribute values) of the original data sets. In this paper we present detailed results of experiments on one data set, showing that some rule sets induced from incomplete data sets are significantly better than the rule set induced from the original data set, with the significance level of 5%, two-tailed test. Additionally, we discuss criteria for inducing better rules by increasing incompleteness and present graphs for some well-known data sets.