Abstract
We describe an experimental study of pruning methods for decision tree classifiers when the goal is minimizing loss rather than error. In addition to two common methods for error minimization, CART's cost-complexity pruning and C4.5's error-based pruning, we study the extension of cost-complexity pruning to loss and one pruning variant based on the Laplace correction. We perform an empirical comparison of these methods and evaluate them with respect to loss. We found that applying the Laplace correction to estimate the probability distributions at the leaves was beneficial to all pruning methods. Unlike in error minimization, and somewhat surprisingly, performing no pruning led to results that were on par with other methods in terms of the evaluation criteria. The main advantage of pruning was in the reduction of the decision tree size, sometimes by a factor of ten. While no method dominated others on all datasets, even for the same domain different pruning mechanisms are better for different loss matrices.
Chapter PDF
Keywords
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.
References
Bradford, J. P., Kunz, C., Kohavi, R., Brunk, C. & Brodley, C. E. (1998), Pruning decision trees with misclassification costs (long). http://robotics.stanford.edu/≈ronnyk/prune-long.ps.gz.
Breiman, L., Friedman, J. H., Olshen, R. A. & Stone, C. J. (1984), Classification and Regression Trees, Wadsworth International Group.
Cestnik, B. (1990), Estimating probabilities: A crucial task in machine learning, in L. C. Aiello, ed., ‘Proceedings of the ninth European Conference on Artificial Intelligence', pp. 147–149.
Draper, B. A., Brodley, C. E. & Utgoff, P. E. (1994), ‘Goal-directed classification using linear machine decision trees', IEEE Transactions on Pattern Analysis and Machine Intelligence 16(9), 888–893.
Good, I. J. (1965), The Estimation of Probabilities: An Essay on Modern Bayesian Methods, M.I.T. Press.
Kohavi, R., Sommerfield, D. & Dougherty, J. (1996), Data mining using MCC++: A machine learning library in C++, in ‘Tools with Artificial Intelligence', IEEE Computer Society Press, pp. 234–245. http://www.sgi.com/Technology/mlc.
Merz, C. J. & Murphy, P. M. (1997), UCI repository of machine learning databases. http://www.ics.uci.edu/≈mlearn/MLRepository.html.
Oates, T. & Jansen, D. (1997), The effects of training set size on decision tree complexity, in D. Fisher, ed., ‘Machine Learning: Proceedings of the Fourteenth International Conference', Morgan Kaufmann, pp. 254–262.
Pazzani, M., Merz, C., Murphy, P., Ali, K., Hume, T. & Brunk, C. (1994), Reducing misclassification costs, in ‘Machine Learning: Proceedings of the Eleventh International Conference', Morgan Kaufmann.
Quinlan, J. R. (1993), C4.5: Programs for Machine Learning, Morgan Kaufmann, San Mateo, California.
Turney, P. (1997), Cost-sensitive learning. http://ai.iit.nrc.ca/bibliographies/cost-sensitive.html.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1998 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Bradford, J.P., Kunz, C., Kohavi, R., Brunk, C., Brodley, C.E. (1998). Pruning decision trees with misclassification costs. In: Nédellec, C., Rouveirol, C. (eds) Machine Learning: ECML-98. ECML 1998. Lecture Notes in Computer Science, vol 1398. Springer, Berlin, Heidelberg. https://doi.org/10.1007/BFb0026682
Download citation
DOI: https://doi.org/10.1007/BFb0026682
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-64417-0
Online ISBN: 978-3-540-69781-7
eBook Packages: Springer Book Archive