Levelwise Search and Pruning Strategies for First-Order HypothesisSpaces

  • Authors:
  • Irene Weber

  • Affiliations:
  • Institut für Informatik, Universität Stuttgart, Breitwiesenstr. 20–22, 70565 Stuttgart, Germany. irene.weber@informatik.uni-stuttgart.de

  • Venue:
  • Journal of Intelligent Information Systems - Special issue on methodologies for intelligent information systems
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

The discovery of interesting patterns inrelational databases is an important data mining task.This paper is concerned with the development of a search algorithm forfirst-order hypothesis spaces adopting an important pruningtechnique (termed subset pruning here) from associationrule mining in a first-order setting. The basic search algorithmis extended by so-called requires and excludes constraintsallowing to declare prior knowledge about the data, suchas mutual exclusion or generalization relationships among attributes,so that it can be exploited for furtherstructuring and restricting the search space. Furthermore, it isillustrated how to process taxonomies and numerical attributes inthe search algorithm.Several task settings using different interestingness criteria andsearch modes with corresponding pruning criteria are described.Three settings serve as test beds for evaluation of theproposed approach. The experimental evaluation shows that theimpact of subset pruning is significant,since it reduces the number of hypothesis evaluations in many cases by about50%. The impact of generalization relationships is shown to beless effective in our experimental set-up.