Sleep Disorder Detection Using Machine Learning

Abstract

Sleep disorders significantlyimpactqualityof life and overall health. This paperpresentsanensemble machine learning approach for accurate sleep disorder classification using demographic and health-related features. We implement and compare five models: Decision Tree, Logistic Regression, XGBoost, Support Vector Machine (SVM), and a Stacking Classifier combining SVM and XGBoost. The dataset undergoes comprehensive preprocessing including label encoding, SMOTE oversampling for class imbalance, and feature standardization. Experimental results demonstrate that XGBoost achieves the highest accuracy (94.95%), followed by the Stacking Classifier (93.43%), SVM (91.92%), Logistic Regression (90.40%), and Decision Tree (87.37%). The proposed approach showssignificant promise for clinical decisionsupportsystems in sleep medicine.

Introduction

Sleep disorders affect nearly 45% of the global population, but diagnosing them accurately is challenging due to complex physiological and lifestyle factors. Traditional diagnostic methods like polysomnography are costly and inaccessible. This study proposes a machine learning framework using a novel stacked ensemble approach that achieves 93.43% accuracy in classifying sleep disorders, outperforming traditional models.

The research addresses key challenges such as imbalanced datasets and feature scaling by applying SMOTE for oversampling and StandardScaler for normalization. Five classifiers were tested: Decision Tree (87.37%), Logistic Regression (90.40%), Support Vector Machine (91.92%), XGBoost (94.95%), and a stacked ensemble combining the best models, achieving 93.43% accuracy.

This approach surpasses previous studies by improving classification performance and handling real-world medical data issues like missing values and heterogeneity. The methodology supports clinical deployment, potentially reducing diagnostic costs and improving early detection of sleep disorders.

The study also reviews related work, highlighting improvements over prior machine learning models and emphasizing the benefits of ensemble methods in medical diagnostics. The dataset included 374 patient records with demographic, physiological, and lifestyle features, processed through label encoding, oversampling, and normalization before model training and evaluation.

In summary, the study presents a practical, high-accuracy machine learning system for sleep disorder classification with clear potential for clinical application.

Conclusion

This study successfully developed a robust machine learning framework for sleep disorder classification, leveraging a stacked ensemble approach to achieve 93.43% accuracy, outperforming individual models including Decision Tree (87.37%), Logistic Regression (90.40%),SVM(91.92%),andXGBoost(94.95%).The methodology incorporated critical preprocessing steps such as Label Encoding for categorical variables, SMOTE-based class balancing, and feature standardization using StandardScaler. The final stacked modelcombinedthestrengthsofSVM(withRBFkernel) and XGBoost (optimized with n_estimators=300 and learning_rate=0.03), using Logistic Regression as a meta-classifier. This approach not only demonstrated superior generalization but also addressed common challenges in medical datasets, such as class imbalance and feature scaling. The results validate the potential of ensemblelearninginclinicaldecisionsupport systemsfor sleep disorder diagnosis, offering a reliable alternative to traditional diagnostic methods

References

[1] Puneeth Kumar Yadav, Uday Kumar Singh. (2024). Sleep Disorder Detection Using Meachine Learning Method. Researchgate [2] LuanaConte , Giorgio De Nunzio , Francesco Giombi Roberto Lupo , Caterina Arigliani (2024), Machine Learning Models to Enhance the Berlin Questionnaire Detection of Obstructive Sleep Apnea in at-Risk Patients, MDPI [3] NattawatSoontreekulpong, Phairot Autthasan, ThitikornKaewlee ,WansirWorawitayanonTanutKhan , Biswas, S. K., &Chunka , (2024) ,ResSense: A Modern Pipeline for Respiratory Arousal Detection Using PPG Signals ,IEEE Xplore [4] Khan, A., Biswas, S. K., &Chunka, C. (2023). ESAD: Expert System for Apnea Detection Using Enhanced DWT Feature Extraction and Machine Learning Algorithms. IEEE ICCCNT. [5] Ashima Arya1, Ramneet Kaur ,DeepamGoyal , Deepali Gupta , Sapna Juneja ,(2023) , Insomnia Detection using Machine Learning Classifiers , IEEE ICCMA [6] Murugesan, M., Abirami, V., Iswarya, M., & Kaviya, N. (2023). Sleeping Abnormalities Detection using Deep Learning Techniques. IEEE ICEARS. [7] AbhyudayaBatabyal,Vinayak Singh, Mahendra Kumar Gourisaria, Himansu Das ,(2022), sleep stress level classification through machine learning algorithms,OCIT [8] PRAGATITRIPATHI ,MDBELALBINHEYAT ,M.A. ANSARI, TAPAN KUMAR GANDHI, RAJAT MEHROTRA ,FAIJANAKHTAR,(2022), Ensemble Computational Intelligent for Insomnia Sleep Stage Detection via the Sleep ECG Signal, IEEEA [9] Guo, Y., Yin, J., Mao, H. X., & Mao, Z. H. (2022). Cluster-Then-Label Strategy for Sleep Detection Using EEG. IEEE ICDSCA. [10] Karunakaran, M., Balusamy, J., & Selvaraj, K. (2022). Machine Learning Models based Mental Health Detection. IEEE ICIC

Copyright

Copyright © 2025 B. Hanumanthu Rao, Challa Venkata Srinadh, Yanamala Lakshmi Prashanthi, Immadi Karthik, Thirdala Prabhu Kiran. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.