A Novel Estimation of the Regularization Parameter for ε-SVM
This paper presents a novel way of estimating the regularization parameter C in regression ε-SVM. The proposed estimation method is based on the calculation of maximum values of the generalization and error loss function terms, present in the objective function of the SVM definition. Assuming that both terms must be optimized in approximately equal conditions in the objective function, we propose to estimate C as a comparison of the new model based on maximums and the standard SVM model. The performance of our approach is shown in terms of SVM training time and test error in several regression problems from well known standard repositories.
KeywordsSupport Vector Machine Loss Function Regularization Parameter Support Vector Regression Support Vector Machine Model
Unable to display preview. Download preview PDF.
- 1.Vapnik, V.N.: Statistical Learning Theory. In: Adaptive and Learning Systems for Signal Processing, Communications and Control. J. Wiley & Sons, Chichester (1998)Google Scholar
- 4.Smola, A.J., Schölkopf, B.: A tutorial on support vector regression. Statistics and Computing (1998)Google Scholar
- 6.Momma, M., Bennett, K.P.: A pattern search method for model selection of support vector regression. In: Proc. of the SIAM International Conference on Data Mining (2002)Google Scholar
- 8.Chang, C.-C., Lin, C.-J.: LIBSVM: a library for support vector machines (2001), http://www.csie.ntu.edu.tw/~cjlin/libsvm
- 9.Asuncion, A., Newman, D.J.: UCI Machine Learning Repository, Irvine, CA: University of California, School of Information and Computer Science (2007), http://www.ics.uci.edu/~mlearn/MLRepository.html
- 10.StatLib DataSets Archive, http://lib.stat.cmu.edu/datasets