Evidential Logistic Regression for Binary SVM Classifier Calibration

详细信息    查看全文

• 作者：Philippe Xu (20)
  Franck Davoine (20) (21)
  Thierry Denoeux (20)
  
• 关键词：Classifier calibration ; theory of belief functions ; Dempster ; Shafer theory ; support vector machines ; logistic regression
• 刊名：Lecture Notes in Computer Science
• 出版年：2014
• 出版时间：2014
• 年：2014
• 卷：8764
• 期：1
• 页码：49-57
• 全文大小：220 KB
• 参考文献：1. Dempster, A.: The Dempster-Shafer calculus for statisticians. International Journal of Approximate Reasoning聽48(2), 365鈥?77 (2008) CrossRef
  2. Den艙ux, T.: A k-nearest neighbor classification rule based on dempster-shafer theory. IEEE Transactions on Systems, Man and Cybernetics聽25(5), 804鈥?13 (1995) CrossRef
  3. Den艙ux, T.: A neural network classifier based on dempster-shafer theory. IEEE Transactions on Systems, Man and Cybernetics聽30(2), 131鈥?50 (2000) CrossRef
  4. Den艙ux, T.: Likelihood-based belief function: justification and some extensions to low-quality data. International Journal of Approximate Reasoning (in Press, 2014), http://dx.doi.org/10.1016/j.ijar.2013.06.007 , doi:10.1016/j.ijar.2013.06.007
  5. Dietterich, T.G.: Approximate statistical tests for comparing supervised classification algorithms. Neural Computation聽10(7), 1895鈥?923 (1998) CrossRef
  6. Kanjanatarakul, O., Sriboonchitta, S., Den艙ux, T.: Forecasting using belief functions: an application to marketing econometrics. International Journal of Approximate Reasoning聽55(5), 1113鈥?128 (2014) CrossRef
  7. Niculescu-Mizil, A., Caruana, R.: Predicting good probabilities with supervised learning. In: Proceedings of the 22nd International Conference on Machine Learning, Bonn, Germany, pp. 625鈥?32 (2005)
  8. Platt, J.C.: Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods. In: Smola, A.J., Bartlett, P., Sch枚lkopf, B., Schuurmans, D. (eds.) Advances in Large-Margin Classifiers, pp. 61鈥?4. MIT Press (1999)
  9. Quost, B., Den艙ux, T., Masson, M.H.: Pairwise classifier combination using belief functions. Pattern Recognition Letters聽28(5), 644鈥?53 (2007) CrossRef
  10. Shafer, G.: A mathematical theory of evidence. Princeton University Press, Princeton (1976)
  11. Sutton-Charani, N., Destercke, S., Den艙ux, T.: Classification trees based on belief functions. In: Denoeux, T., Masson, M.H. (eds.) Belief Functions: Theory and Applications. AISC, vol.聽164, pp. 77鈥?4. Springer, Heidelberg (2012) CrossRef
  12. Wu, T.F., Lin, C.J., Weng, R.C.: Probability estimates for multi-class classification by pairwise coupling. Journal of Machine Learning Research聽5, 975鈥?005 (2004)
  13. Zadrozny, B., Elkan, C.: Obtaining calibrated probability estimates from decision trees and naive Bayesian classifiers. In: Proceedings of the 18th International Conference on Machine Learning, Williamstown, USA, pp. 609鈥?16 (2001)
  14. Zadrozny, B., Elkan, C.: Transforming classifier scores into accurate multiclass probability estimates. In: Proceedings of the 8th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, New York, USA, pp. 694鈥?99 (2002)
  
• 作者单位：Philippe Xu (20)
  Franck Davoine (20) (21)
  Thierry Denoeux (20)
  
  20. UMR CNRS 7253, Heudiasyc, Universit茅 de Technologie de Compi猫gne, France
  21. CNRS, LIAMA, Beijing, P.聽R. China
  
• ISSN：1611-3349

文摘

The theory of belief functions has been successfully used in many classification tasks. It is especially useful when combining multiple classifiers and when dealing with high uncertainty. Many classification approaches such as k-nearest neighbors, neural network or decision trees have been formulated with belief functions. In this paper, we propose an evidential calibration method that transforms the output of a classifier into a belief function. The calibration, which is based on logistic regression, is computed from a likelihood-based belief function. The uncertainty of the calibration step depends on the number of training samples and is encoded within a belief function. We apply our method to the calibration and combination of several SVM classifiers trained with different amounts of data.