Abstract
This paper provides an overview of calibration methods for supervised classification learners. Calibration means a scaling of classifier scores into the probability space. Such a probabilistic classifier output is especially useful if the classification output is used for post-processing. The calibraters are compared by using 10-fold cross-validation according to their performance on SVM and CART outputs for four different two-class data sets.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
BENNETT, P. (2003): Using Asymmetric Distributions to Improve Text Classifier Probability Estimates: A Comparison of New and Standard Parametric Methods. Techn. Report CMU-CS-02-126, Carnegie Mellon, School of Computer Science.
GARCZAREK, U. (2002): Classification Rules in Standardized Partition Spaces. [http://eldorado.uni-dortmund.de:8080/FB5/ls7/forschung/2002/Garczarek]. Dissertation, Universität Dortmund.
PLATT, J. (1999): Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods. In: A. Smola, P. Bartlett, B. Schoelkopf and D. Schuurmans (Eds.): Advances in Large Margin Classiers. MIT Press, Cambridge, 61–74.
ZADROZNY, B. and ELKAN, C. (2002): Transforming Classifier Scores into Accurate Multiclass Probability Estimates. In: Proceedings of the 8th Internat. Conference on Knowledge Discovery and Data Mining. ACM Press, Edmonton, 694–699.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2007 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Gebel, M., Weihs, C. (2007). Calibrating Classifier Scores into Probabilities. In: Decker, R., Lenz, H.J. (eds) Advances in Data Analysis. Studies in Classification, Data Analysis, and Knowledge Organization. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-70981-7_17
Download citation
DOI: https://doi.org/10.1007/978-3-540-70981-7_17
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-70980-0
Online ISBN: 978-3-540-70981-7
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)