Automatic Sleep Stage Classification using Weighted ELM and PSO on Imbalanced Data from Single Lead ECG

Sleep stage classification is one of important aspects in sleep studies, which can give clinical information for diagnosing sleep disorder and measuring sleep quality. Due to the difference in sleep stage proportion for every person, the collected sleep stage data are imbalanced naturally, which can lead to high probability of misclassification. Various learning method has been developed to classify sleep stage based on electrocardiogram (ECG) signal. However, to the best of our knowledge, there are no researches which consider the imbalanced dataset problem for sleep stage classification. In this research, a classification model of sleep stage based on ECG signal was developed using Weighted Extreme Machine Learning (WELM) to deal with imbalanced learning dataset and Particle Swarm Optimization (PSO) for feature selection. The research will use the MIT-BIH Poly sorrmographic Database, which contains 10154 sleep stage annotated ECG data which consist of 17.79%, 38.28%, 4.76%, 1.78%, 6.89%, and 30.5% data of NREM1, NREM2, NREM3, NREM4, REM, and awake stage respectively. From each ECG record, a total of 18 features were extracted and the feature selection process resulted in 10 features which highly affect the sleep stage classification. The proposed model successfully obtained a mean accuracy of 78,78% for REM, NREM and Wake stage classification and 73.09% for Light Sleep, Deep Sleep, REM, and Wake stage classification. (C) 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the scientific committee of the 4th International Conference on Computer Science and Computational Intelligence 2019.