An Ensemble Feature Selection Approach to Identify Relevant Features from EEG Signals

Identifying relevant data to support the automatic analysis of electroencephalograms (EEG) has become a challenge. Although there are many proposals diagnosis neurological pathologies, current challenge is improve reliability tools classify or detect abnormalities. In this study, we used an ensemble feature selection approach integrate advantages several algorithms identification characteristics with high power differentiation in classification normal and abnormal EEG signals. Discrimination was evaluated using classifiers, i.e., decision tree, logistic regression, random forest, Support Vecctor Machine (SVM); furthermore, performance assessed by accuracy, specificity, sensitivity metrics. The evaluation results showed that Ensemble Feature Selection (EFS) helpful tool select features from EEGs. Thus, stability calculated for EFS method proposed almost perfect most cases evaluated. Moreover, classifiers evidenced models improved when trained approach’s features. addition, classifier epileptiform events built selected achieved sensitivity, specificity 97.64%, 96.78%, 97.95%, respectively; finally, reliable subset detector equal greater than values reported literature.