Two-Class with Oversampling Versus One-Class Classification for Microarray Datasets

Microarray datasets are a challenge for classical computational techniques because of the large dimensionality of their feature space front to a reduced number of samples, besides they usually present unbalanced classes. Thanks to this unbalanced situation, in a previous research, the superiority of one-class classification for handling microarray datasets was proved. This paper presents a new study that tries to improve the behavior of the traditional techniques, specifically Support Vector Machines, by considering oversampling techniques. The experimental results achieved demonstrate that despite inclusion of these methods the performance of classical classifiers still remains below one-class approach.