Tree-Based Machine Learning Techniques for Automated Human Sleep Stage Classification

Creative Commons License


Traitement du Signal, vol.40, no.4, pp.1385-1400, 2023 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 40 Issue: 4
  • Publication Date: 2023
  • Doi Number: 10.18280/ts.400408
  • Journal Name: Traitement du Signal
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, PASCAL, Business Source Elite, Business Source Premier, Compendex, zbMATH
  • Page Numbers: pp.1385-1400
  • Keywords: machine learning, multi-channel data, polysomnography (PSG), sleep stage scoring
  • Kayseri University Affiliated: Yes


Background: Sleep disorders pose significant health risks, necessitating accurate diagnostics. The analysis of polysomnographic data and subsequent sleep stage classification by medical professionals are crucial in diagnosing these disorders. The application of artificial intelligence (AI)-based systems for automated sleep stage classification has gained significant momentum recently. Methodology: In this study, we introduce a machine learning model designed for high-accuracy, automated sleep stage classification. We utilized a dataset consisting of polysomnographic data from 50 individuals, obtained from the Yozgat Bozok University Sleep Center. A variety of classifiers, including Extra Tree, Decision Tree, Random Forest, Ada Boost, and Gradient Boost, were tested. Sleep stages were classified into three categories: Wakefulness (WK), Rapid Eye Movement (REM), and Non-Rapid Eye Movement (N-REM). Results: The overall classification accuracies were 95.4%, 95%, and 92% for three distinct classifiers, respectively, with the highest accuracy reaching 98.8%. Comparison with Existing Methods: This study distinguishes itself from comparable sleep stage-scoring research by utilizing a unique dataset, and by incorporating data from 16 channels, which contributes to the achieved accuracy. Conclusion: The machine learning model trained with a unique dataset demonstrated high classification success in the automated scoring of sleep stages. This research underscores the potential of machine learning techniques in improving sleep disorder diagnostics.