| dc.description.abstract | This study advances the field of sleep stage classification by enhancing the
performance of a deep learning model, AttnSleep. The methodology involved
substituting the standard Multi-resolution Convolutional Neural Network with a
Multiscale Multiperiod Convolutional Neural Network for feature extraction. Singlechannel EEG data from the Sleep-EDF-20 database served as the basis for the
experiments. The primary research problem revolves around improving the
misclassification of sleep stages. Results showed a rise in MF1, suggesting improved
performance with imbalanced datasets. Despite continued misclassification with the N2
stage, there was notable progress in N1 classification. The research outcomes could aid
clinicians in diagnosing and treating sleep disorders more effectively and provide
patients with more accessible, personalized assessments. Future research should focus
on utilizing more extensive, diverse datasets, analyzing misclassification patterns,
enhancing transfer learning, and considering ensemble learning strategies. | en_US |
