An improved simulated annealing genetic algorithm of EEG feature selection in sleep stage

In order to establish a more reliable and robust EEG model in sleep stages, the reasonable choice of modeling parameters is necessary. The function of this step is to select a subset of d features from a set of D features based on some optimization criterion, and provide the most optimal input features of classification. In the present study, an improved simulated annealing genetic algorithm (ISAGA) was proposed. 25 feature parameters were extracted from the sleep EEG in MIT-BIH polysomnography database. The feature selection results demonstrated that ISAGA can get a higher classification accuracy with fewer feature number than the correlation coefficient algorithm (CCA), genetic algorithm (GA), adaptive genetic algorithm (AGA) and simulated annealing genetic algorithm (SAGA). Compared to using all the features in sleep staging, the classification accuracy of ISAGA with optimal features is about 92.00%, which improved about 4.83%.