CLASS IMBALANCE LEARNING METHODS FOR SUPPORT VECTOR MACHINES - Imbalanced Learning: Foundations, Algorithms, and Applications - page 94

Following the aforementioned methods of assigning membership values for
positive and negative training data points, several FSVM-CIL settings have been
defined in [40]. These methods have been validated on 10 real-world imbal-
anced datasets representing a variety of domains, complexities, and imbalanced
ratios, which are highly likely to contain noisy examples and outliers. FSVM-
CIL settings have resulted in better classification results on these datasets than the
existing class imbalance learning methods applied for standard SVMs, namely
random oversampling, random undersampling, SMOTE, DEC, and zSVM meth-
ods. Batuwita and Palade [40] pointed out that better performance of FSVM-CIL
method is due to its capability to handle outliers and noise in these datasets in
addition to the class imbalance problem.
5.5.7 Hybrid Methods
There exist methods that have used the combination of both external and internal
methods to solve the class imbalance problem for SVMs. The hybrid kernel
machine ensemble (HKME) method [43] combines a standard binary SVM and
a one-class SVM classifier to solve the problem of class imbalance. Akbani et
al. [10] has combined the SMOTE algorithm with the DEC method for SVMs
for imbalanced dataset learning and shown to have better performance than the
use of either of these methods alone.
5.6 SUMMARY
This chapter aimed to review the existing imbalance learning methods developed
for SVMs. These methods have been developed as data pre-processing methods
or algorithmic improvements. As pointed out in the literature, the class imbalance
learning method giving the optimal solution is often dataset dependent. Therefore,
it is worth applying several of these available external and internal methods and
compare the performances, when training an SVM model on an imbalanced
dataset.
REFERENCES
1. V. Vapnik, The Nature of Statistical Learning Theory . New York: Springer-Verlag,
Inc., 1995.
2. C. Cortes and V. Vapnik, “Support-vector networks,” Machine Learning , vol. 20, no.
3, pp. 273-297, 1995.
3. B. Boser, I. Guyon, and V. Vapnik, “A training algorithm for optimal margin classi-
fiers,” in Proceedings of the 5th Annual ACM Workshop on Computational Learning
Theory (Pittsburgh, PA, USA) , pp. 144-152, ACM Press, 1992.
4. N. Cristianinio and J. Shawe-Taylor, An Introduction to Support Vector Machines
and Other Kernel-Based Learning Methods . Cambridge: Cambridge University Press,
2000.