Advertisement

Theoretical analysis of case retrieval method based on neighborhood of a new problem

  • Seishi Okamoto
  • Nobuhiro Yugami
Scientific Papers Indexing And Retrieval
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1266)

Abstract

The retrieval of similar cases is often performed by using the neighborhood of a new problem. The neighborhood is usually denned by a certain fixed number of most similar cases (k nearest neighbors) to the problem. This paper deals with an alternative definition of neighborhood that comprises the cases within a certain distance, d, from the problem. We present an average-case analysis of a classifier, the d-nearest neighborhood method (d-NNh), that retrieves cases in this neighborhood and predicts their majority class as the class of the problem. Our analysis deals with m-of-n/l target concepts, and handles three types of noise. We formally compute the expected classification accuracy of d-NNh, then we explore the predicted behavior of d-NNh. By combining this exploration for d-NNh and one for k-nearest neighbor method (k-NN) in our previous study, we compare the predicted behavior of each in noisy domains. Our formal analysis is supported with Monte Carlo simulations.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Aha, D., Kibler, D., and Albert, M. Instance-Based Learning Algorithms. Machine Learning, 6, (1991) 37–66.Google Scholar
  2. 2.
    Albert, M. and Aha, D. Analyses of Instance-Based Learning Algorithms. In Proceedings of AAAI-91, (1991) 553–558. AAAI Press/MIT Press.Google Scholar
  3. 3.
    Cover. T. and Hart, P. Nearest Neighbor Pattern Classification. IEEE Transactions on Information Theory. 13(1). (1967) 21–27.Google Scholar
  4. 4.
    Creecy, H., Masand, M., Smith, J., and Waltz, D. Trading Mips and Memory for Knowledge Engineering. Communications of the A CM, 35(8), (1992) 48–63.Google Scholar
  5. 5.
    Drakopoulos, J. Bounds on the Classification Error of the Nearest Neighbor Rule. In Proceedings of ICML-95, (1995) 203–208. Morgan Kaufmann.Google Scholar
  6. 6.
    Langley. P. and Iba, W. Average-Case Analysis of a Nearest Neighbor Algorithm. In Proceedings of IJCAI-93, (1993) 889–894. Morgan Kaufmann.Google Scholar
  7. 7.
    Murphy, P. and Pazzani, M. ID2-of-3: Constructive Induction of M-of-N Concepts for Discriminators in Decision Trees. In Proceedings of IWML-91, (1991) 183–187. Morgan Kaufmann.Google Scholar
  8. 8.
    O'Callaghan, J. P. An Alternative Definition for Neighborhood of a Point, IEEE Transactions on Computers, 24(11), (1975) 1121–1125.Google Scholar
  9. 9.
    Okamoto, S. and Satoh, K. An Average-Case Anaysis of k-Nearest Neighbor Cassifier. In Proceedings of ICCBR-95 (Veoso, M. and Aamodt, A. Eds., LNAI, 1010), (1995) 243–264. Springer-Verag.Google Scholar
  10. 10.
    Okamoto, S. and Yugami, N. Theoretica Anaysis of the Nearest Neighbor Cassifier in Noisy Domains. In Proceedings of ICML-96 (1996) 355–363. Morgan Kaufmann.Google Scholar
  11. 11.
    Okamoto, S. and Yugami, N. An Average-Case Anaysis of the k-Nearest Neighbor Cassifier for Noisy Domains. In Proceedings of IJCAI-97, (1997) to appear. Morgan Kaufmann.Google Scholar
  12. 12.
    Pazzani, M. and Sarrett, W. A Framework for Average Case Anaysis of Conjunctive Learning Agorithms. Machine Learning, 9 (1992) 349–372.Google Scholar
  13. 13.
    Wettschereck, D. and Aha, D. Weighting Features. In Proceedings of ICCBR-95 (Veoso, M. and Aamodt, A. Eds., LNAI, 1010), (1995) 347–358. Springer-Verag.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1997

Authors and Affiliations

  • Seishi Okamoto
    • 1
  • Nobuhiro Yugami
    • 1
  1. 1.Fujitsu Laboratories LimitedFukuokaJapan

Personalised recommendations