Conceptual Neighborhoods for Retrieval in Case-Based Reasoning

  • Authors:
  • Ben G. Weber;Michael Mateas

  • Affiliations:
  • University of California, Santa Cruz, Santa Cruz, USA CA 95064;University of California, Santa Cruz, Santa Cruz, USA CA 95064

  • Venue:
  • ICCBR '09 Proceedings of the 8th International Conference on Case-Based Reasoning: Case-Based Reasoning Research and Development
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present a case-based reasoning technique based on conceptual neighborhoods of cases. The system applies domain knowledge to the case retrieval process in the form of recall and generalize methods. Recall methods utilize domain specific preconditions and perform exact matching, while generalize methods apply transformations that generalize features in queries. The system uses a similarity function based on edit distances, where an edit distance considers only a subset of the features. This retrieval strategy enables the system to locate conceptually similar cases within the feature space. We demonstrate the performance of this approach by applying it to build-order selection in a real-time strategy game. Our results show that the system outperforms nearest neighbor retrieval when enforcing imperfect information in a real-time strategy game.