An ART-Based Hybrid Network for Medical Pattern Classification Tasks with Missing Data

  • Chee Peng Lim
  • Mei Ming Kuan
  • Robert F. Harrison
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 2774)

Abstract

In this paper, a hybrid neural network capable of incremental learning and classification of patterns with missing features is proposed. Fuzzy ARTMAP (FAM) is employed as the constituting network for pattern classification while a number of Fuzzy c-Means (FCM) clustering algorithms are used for undertaking the incomplete data problem. To handle training samples with missing features, FAM is first trained with complete samples. Training samples containing missing features are then presented, and the missing values are estimated and replaced by using two FCM-based algorithms. After that, network training is conducted again with all the complete and estimated samples. To handle test samples with missing features, a non-substitution FCM-based approach is employed to yield a predicted output rapidly. A real medical database is used to evaluate, empirically, the performance of the hybrid FAM-FCM network. The results are analysed, discussed, and quantified statistically with the bootstrap method.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Carpenter, G.A., Grossberg, S.: A Massively Parallel Architecture for a Selforganizing Neural Pattern Recognition Machine. Computer Vision Graphics and Image Processing 37, 54–115 (1987)CrossRefGoogle Scholar
  2. 2.
    Carpenter, G.A., Grossberg, S., Markuzon, N., Reynolds, J.H., Rosen, D.B.: Fuzzy ARTMAP: A Neural Network Architecture for Incremental Supervised Learning of Analog Multidimensional Maps. IEEE Trans. Neural Networks 3, 698–712 (1992)CrossRefGoogle Scholar
  3. 3.
    Bezdek, J.C.: Pattern Recognition with Fuzzy Objective Function Algorithms. Plenum Press, New York (1981)MATHGoogle Scholar
  4. 4.
    Cano Izquierdo, J., Dimitriadis, Y., Gómez Sánchez, E., López-Coronado, J.: Learning from Noisy Information in FasArt and FasBack Neuro-fuzzy Systems. Neural Networks 14, 407–425 (2001)CrossRefGoogle Scholar
  5. 5.
    Carpenter, G.A., Ross, W.: ART-EMAP: A Neural Network Architecture for Object Recognition by Evidence Accumulation. IEEE Trans. Neural Networks 6, 805–818 (1995)CrossRefGoogle Scholar
  6. 6.
    Williamson, J.R.: Gaussian ARTMAP: A Neural Network for Fast Incremental Learning of Noisy Multidimensional Maps. Neural Networks 9, 881–897 (1996)CrossRefGoogle Scholar
  7. 7.
    Carpenter, G.A., Milenova, B.L., Noeske, B.W.: Distributed ARTMAP: A Neural Network for Fast Distributed Supervised Learning. Neural Networks 11, 793–813 (1998)CrossRefGoogle Scholar
  8. 8.
    Granger, E., Rubin, M.A., Grossberg, S., Lavoie, P.: Classification of Incomplete Data using the Fuzzy ARTMAP Neural Network. In: Proc. IEEE-INNSENNS Int. Joint Conf. on Neural Networks, vol. 6, pp. 35–40 (2001)Google Scholar
  9. 9.
    Hathaway, R.J., Bezdek, J.C.: Fuzzy c-Means Clustering with Incomplete Data. IEEE Trans. Syst. 31, 735–744 (2001)Google Scholar
  10. 10.
    Hathaway, R.J., Hu, Y., Bezdek, J.C.: Local Convergence of Tri-level Alternating Optimisation. Neural, Parallel, Sci. Computat. 9, 19–28 (2001)MATHMathSciNetGoogle Scholar
  11. 11.
    Lim, C.P., Harrison, R.F.: An Incremental Adaptive Network for On-Line Supervised Learning and Probability Estimation. Neural Networks 10, 925–939 (1997)CrossRefGoogle Scholar
  12. 12.
    Efron, B.: Bootstrap Methods: Another Look at the Jackknife. The Annals of Statistics 7, 1–26 (1979)MATHCrossRefMathSciNetGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2003

Authors and Affiliations

  • Chee Peng Lim
    • 1
  • Mei Ming Kuan
    • 1
  • Robert F. Harrison
    • 2
  1. 1.School of Electrical & Electronic EngineeringUniversity of Science MalaysiaNibong Tebal, PenangMalaysia
  2. 2.Department of Automatic Control and Systems EngineeringThe University of SheffieldSheffieldUnited Kingdom

Personalised recommendations