| Autor: |
Mounia, Hendel, Abdelkader, Benyettou, Fatiha, Hendel |
| Předmět: |
|
| Zdroj: |
International Review on Computers & Software; Jan2016, Vol. 11 Issue 1, p49-55, 7p |
| Abstrakt: |
The support vector machines (SVM) are a powerful approach which has been applied in many difficult applications. However, they are principally designed for binary classification problems, and usually their extension for multiclass problems involves only bi-class SVM. In this study, a recently developed machine named the quadratic loss multi-class SVM (M-SVM2), which considers all classes simultaneously, is proposed to classify five different arrhythmia, in addition to normal beat. The M-SVM2 is compared with both decomposition methods involving 2-norm binary SVM (l2- SVM) based on 1-against-1, and 1-against-all approach. The proposed model achieved an average accuracy of (99.73%), which was better compared to the other implemented classifiers. On the other hand, the study showed that post-processing the outputs of the M-SVM2 in terms of probability can significantly improve the classification decision. This results have shown the effectiveness of the proposed approach to enhance the performance of Electrocardiogram (EEG) classification methods. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
