An Ensemble Model for Sleep Stages Classification

One of the main parts of health is the quality of sleep. Sleep disorders can be diagnosed using a standard sleep test called polysomnography. Sleep staging is a task in the field of sleep study that determines sleep cycles. In recent years, machine learning techniques are used to classify sleep stages using biological signals derived from PSGs. In this study, we propose an ensemble machine learning model for classifying sleep stage. We use nine biological signals from the SHHSI dataset, including two-channel EEG, two-channel EOG, ECG, EMG, abdominal, thorax, and airflow signals. Then we extract different features such as RRI and RPE from the ECG signals and frequency features from EEG signals. Finally, we develop an ensemble model using Light Gradient Boost (LGB) and eXtreme Gradient Boost (XGB) algorithms. In the end, we evaluate the proposed ensemble method using different metrics and compare its performance with other state-of-the-art techniques. The results of the proposed method show that it achieves an overall accuracy of 0.8951 in the five-class classification of sleep stages including Wake, Nl, N2, N3, and REM.