A GA-SVM Model for Sleep/Wake Classification Using Single-Channel EEG Signals

The primary focus of this study is the development of an SVM-based method for classifying sleep/wake states using single-channel EEG signals. Through the optimization of the SVM model via a genetic algorithm (GA), optimal parameters (C=83.6932, gamma=0.0050) were obtained from the fitness curve. Building the GA-SVM model upon these parameters resulted in an impressive accuracy of 96.94% and a Cohen's kappa statistic of 0.9389, highlighting the model's exceptional classification performance. This model offers a reliable approach to precisely identify moments of sleep onset and awakening during subsequent sleep monitoring endeavors.