Development of Efficient Ensemble Model based on Stacking Learning for Automated Sleep Staging

Sleep plays a vital role in maintaining good physical and mental health. Improper sleep causes many critical diseases; sleep-related disorders are significant global challenges. The primary diagnosis step for such disorders is a classification of sleep stages. Compared to traditional complex manual sleep analysis, automated sleep staging methods using single-channel electroencephalography (EEG) signals more practical benefits for analyzing the changes in characteristics in different sleep stages. Many of the existing contributions are based on computer-aided techniques for this task; among these, some of these classification models performed well to improve the classification accuracy results, but the achieved accuracy has not been reached for clinical applications due to improper selection of features. In this work, we consider an effective ensemble learning algorithm for improving the classification accuracy of sleep stages. Previous research works have not performed satisfactory classification performance, ignored some compelling features, and not identified suitable ones. This work extracts different feature extraction to obtain valuable features from EEG signals. Meanwhile, an effective classification process based on ensemble learning is proposed to improve the classification results automatically. The proposed model evaluated eight subjects of the ISRUC-Sleep public dataset, and the model reported accuracy of 97.86% using subgroup-I and 99.02% using sub-group-III data, respectively. Experimental results show that the classification performance of the proposed method is incomparably enhanced compared to the state-of-the-art works. This sleep stage classification framework is expected to assist the medical professional in diagnosing the different types of sleep-related disorders.