Polichotomies on Imbalanced Domains by One-per-Class Compensated Reconstruction Rule

  • Roberto D’Ambrosio
  • Paolo Soda
Part of the Lecture Notes in Computer Science book series (LNCS, volume 7626)


A key issue in machine learning is the ability to cope with recognition problems where one or more classes are under-represented with respect to the others. Indeed, traditional algorithms fail under class imbalanced distribution resulting in low predictive accuracy over the minority classes. While large literature exists on binary imbalanced tasks, few researches exist for multiclass learning. In this respect, we present here a new method for imbalanced multiclass learning within the One-per-Class decomposition framework. Once the multiclass task is divided into several binary tasks, the proposed reconstruction rule discriminates between safe and dangerous classifications. Then, it sets the multiclass label using information on both data distributions and classification reliabilities provided by each binary classifier, lowering the effects of class skew and improving the performance. We favorably compare the proposed reconstruction rule with the standard One-per-Class method on ten datasets using four classifiers.


Minority Class Facial Expression Recognition Class Imbalance Multiclass Problem Decomposition Framework 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


  1. 1.
    Chawla, N.V., et al.: SMOTE: Synthetic minority over-sampling technique. J. of Artificial Intelligence Research 16(3), 321–357 (2002)zbMATHGoogle Scholar
  2. 2.
    Soda, P.: A multi-objective optimisation approach for class-imbalance learning. Pattern Recognition 44, 1801–1810 (2011)zbMATHCrossRefGoogle Scholar
  3. 3.
    Alejo, R., et al.: An empirical study for the multi-class imbalance problem with neural networks. Progress in Pattern Recog., Image Analysis and Applications, 479–486 (2008)Google Scholar
  4. 4.
    Fernández, et al.: Multi-class imbalanced data-sets with linguistic fuzzy rule based classification systems based on pairwise learning. Comp. Intel. for Knowledge-Based Systems Design, 89–98 (2010)Google Scholar
  5. 5.
    Martínez-Estudillo, F.J., et al.: Evolutionary learning by a sensitivity-accuracy approach for multi-class problems. In: IEEE WCCI, pp. 1581–1588 (2008)Google Scholar
  6. 6.
    Soda, P., et al.: Decomposition methods and learning approaches for imbalanced dataset: An experimental integration. In: 20th Int. Conf. on Pattern Recognition, pp. 3117–3120 (2010)Google Scholar
  7. 7.
    Wu, M., et al.: A small sphere and large margin approach for novelty detection using training data with outliers. IEEE Tran. on Pattern Anal. and Machine Intel. 31(11), 2088–2092 (2009)CrossRefGoogle Scholar
  8. 8.
    Allwein, E.L., et al.: Reducing multiclass to binary: a unifying approach for margin classifiers. J. of Machine Learning Research 1, 113–141 (2001)MathSciNetzbMATHGoogle Scholar
  9. 9.
    Dietterich, T.G., et al.: Solving multiclass learning problems via error-correcting output codes. J. of Artificial Intelligence Research 2, 263 (1995)zbMATHGoogle Scholar
  10. 10.
    Fürnkranz, J.: Round robin classification. JMLR 2, 721–747 (2002)zbMATHGoogle Scholar
  11. 11.
    Masulli, F., et al.: Comparing decomposition methods for classication. In: 4th Intl. Conf. on Knowledge-Based Intel. Engineering Systems & Allied Technologies, pp. 788–791 (2000)Google Scholar
  12. 12.
    Rajan, S., et al.: An Empirical Comparison of Hierarchical vs. Two-Level Approaches to Multiclass Problems. In: Roli, F., Kittler, J., Windeatt, T. (eds.) MCS 2004. LNCS, vol. 3077, pp. 283–292. Springer, Heidelberg (2004)CrossRefGoogle Scholar
  13. 13.
    Dietterich, T., et al.: Solving multiclass learning problem via error-correcting output codes. J. of Artificial Intel. Research 2, 263–286 (1995)zbMATHGoogle Scholar
  14. 14.
    Mayoraz, E., et al.: On the decomposition of polychotomies into dichotomies. In: Proc. of the 14th Int. Conf. on Machine Learning, pp. 219–226 (1997)Google Scholar
  15. 15.
    Iannello, G., et al.: On the use of classification reliability for improving performance of the one-per-class decomposition method. Data & Knowledge Eng. 68, 1398–1410 (2009)CrossRefGoogle Scholar
  16. 16.
    Foggia, P., et al.: On Rejecting Unreliably Classified Patterns. In: Haindl, M., Kittler, J., Roli, F. (eds.) MCS 2007. LNCS, vol. 4472, pp. 282–291. Springer, Heidelberg (2007)CrossRefGoogle Scholar
  17. 17.
    Cordella, L.P., et al.: Reliability parameters to improve combination strategies in multi-expert systems. Pattern Analysis & Applications 2(3), 205–214 (1999)CrossRefGoogle Scholar
  18. 18.
    Platt, J.: Probabilistic output for support vector machines and comparisons to regularize likelihood methods. Advanced in Large Margin Classifiers. MIT Press (2000)Google Scholar
  19. 19.
    Kanade, T., et al.: Comprehensive database for facial expression analysis. In: 4h IEEE Int. Conf. on Automatic Face and Gesture Recognition, pp. 46–53 (2000)Google Scholar
  20. 20.
    D’Ambrosio, R., et al.: Automatic Facial Expression Recognition Using Statistical-Like Moments. In: Maino, G., Foresti, G.L. (eds.) ICIAP 2011, Part I. LNCS, vol. 6978, pp. 585–594. Springer, Heidelberg (2011)CrossRefGoogle Scholar
  21. 21.
    Soda, P., et al.: A multiple experts system for classifying fluorescence intensity in antinuclear autoantibodies analysis. Pattern Analysis & Applications 12(3), 215–226 (2009)MathSciNetCrossRefGoogle Scholar
  22. 22.
    Soda, P., et al.: Aggregation of classifiers for staining pattern recognition in antinuclear autoantibodies analysis. IEEE T Inf. Technol. B 13(3), 322–329 (2009)MathSciNetCrossRefGoogle Scholar
  23. 23.
    Frank, A., et al.: UCI machine learning repository (2010)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • Roberto D’Ambrosio
    • 1
  • Paolo Soda
    • 1
  1. 1.Integrated Research CentreUniversitá Campus Bio-Medico of RomeRomeItaly

Personalised recommendations