A Unifying View on Instance Selection

  • Authors:
  • Thomas Reinartz

  • Affiliations:
  • DaimlerChrysler AG, Research and Technology, FT3/KL, P.O. Box 2360, 89013 Ulm, Germany. thomas.reinartz@daimlerchrysler.com

  • Venue:
  • Data Mining and Knowledge Discovery
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we consider instance selection as an important focusing task in the data preparation phase of knowledge discovery and data mining. Focusing generally covers all issues related to data reduction. First of all, we define a broader perspective on focusing tasks, choose instance selection as one particular focusing task, and outline the specification of concrete evaluation criteria to measure success of instance selection approaches. Thereafter, we present a unifying framework that covers existing approaches towards solutions for instance selection as instantiations. We describe specific examples of instantiations of this framework and discuss their strengths and weaknesses. Then, we outline an enhanced framework for instance selection, generic sampling, and summarize example evaluation results for several different instantiations of its implementation. Finally, we conclude with open issues and research challenges for instance selection as well as focusing in general.