Comparative Analysis of Conventional and Ensemble Machine Learning Techniques for Sleep Disorder Classification

Authors : Ceren Kaya

Pages : 132-139

View : 81 | Download : 190

Publication Date : 2025-12-23

Article Type : Research Paper

Abstract :Computer-aided automatic diagnosis systems have become increasingly important in early detection and management of sleep disorders, as they provide rapid analytical capabilities and assist clinicians in making more accurate and consistent decisions. In this study, two prevalent sleep disorders, apnea and insomnia, were classified using four widely utilized machine learning techniques: Naive Bayes, Support Vector Machine, Logistic Regression, and Random Forest. Each classifier offers a unique analytical perspective, thereby contributing to a more comprehensive assessment of diagnostic performance. To further enhance the robustness and reliability of classification results, three ensemble learning algorithms (AdaBoost, Bagging, and Random Subspace) are employed and integrated with base classifiers. The performance of all models was quantitatively assessed using several key evaluation metrics, including accuracy, kappa coefficient, precision and Area Under the Curve, which together provide a holistic view of classification quality. According to the experimental findings, Logistic Regression stands out as the most effective individual classifier, achieving the highest accuracy rate of 94.667%. Moreover, when Logistic Regression was combined with Bagging or Random Subspace ensemble methods, additional improvements were observed across all evaluation criteria, demonstrating the potential of ensemble-based approaches to further strengthen automated sleep disorder diagnosis.
Keywords : Sınıflandırma, Hastalık Tahmini, Topluluk Öğrenmesi, Makine Öğrenmesi, Uyku Bozuklukları

ORIGINAL ARTICLE URL