Sleep Disorder Prediction System Using Machine Learning

Insomnia and sleep apnea result in extreme wellbeing deterioration which requires reliable but economical diagnostic methods. Traditional diagnosis methods known as polysomnographies require considerable resources therefore machine learning methods become necessary. The research checks classification of sleep disorders by evaluating Random Forest and Bagging combined with SVM alongside Logistic Regression. Among them, Bagging achieved the highest accuracy (92.4 %) with precision and recall of 92.4 %, further validated by 5fold cross-validation (91.8 % average accuracy). Hyperparameter tuning via grid search enhanced model performance, with an optimized Random Forest attaining <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 2. 4 \%}$</tex> accuracy and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 2. 7 \%}$</tex> precision. An ablation study identified sleep duration, quality, and BMI as key predictive features. These findings underscore the potential of machine learning, particularly Bagging and Logistic Regression, for scalable, cost-effective sleep disorder diagnostics.