Tree-based EEG Classifiers for Somnolence Detection: Model Search Computational Experiments

Authors: Jan Klaschka
Publication year: 2011
Publication type: Kapitola v monografii
Periodical: Reliability of Driver Car Interactions
Projects: Analysis of negative impacts on driver attention
Bibliographic citation: KLASCHKA, Jan. Tree-based EEG Classifiers for Somnolence Detection: Model Search Computational Experiments. In: NOVÁK, M., ed. Reliability of Driver Car Interactions. Praha: Ústav informatiky AV ČR, v. v. i. a ČVUT, Fakulta dopravní, 2011, S. 80-96. Edice monografií NNW, 9. ISBN 978-80-87136-12-6.
ID: 372954

Abstract

Different strategies of the search for a tree-based EEG classification model aimed at driver's somnolence detection are applied to real data and compared in a computational experiment. Furthermore, a follow-up experimental study of one of the "winning" models from the former experiment is reported. The latter experiment is specifically focused at model search restriction as a method of coping with some confusing phenomena observed when varying parameters of the model.