Classification of EEG Data using Fuzzy k-NN Ensembles
Abstract
Ensemble methods try to improve quality of classification by creating multiple classifiers and aggregating their outputs. In this paper, we present the use of ensemble methods for classification of EEG data from the project "Building Neuroinformation Bases, and Extracting Knowledge from them", within which a possibility of preventing drivers' microsleeps is studied. A multiple feature subset ensemble method is used to improve the quality of classification of a fuzzy k-nearest neighbor classifier. Two different aggregation schemes are used - the mean value aggregation algorithm outperforming the Sugeno fuzzy integral aggregation algorithm