Forecasting Sleep Disorders and Sleep Efficiency Using Machine Learning Techniques

Sleep, being a necessity and not just a luxury, is often ignored to the extent that sleep disorders have become increasingly prevalent in the context of global health. Traditional diagnostic techniques like polysomnography are often inconvenient as they require expensive equipment, clinical experts, and overnight monitoring. Sleep apnea and insomnia are the primary sleep disorders that lead to severe health issues without early diagnosis and treatment. To forecast sleep disorders and predict sleep efficiency aiming to provide a convenient solution this study focuses on utilizing machine learning and deep learning techniques. The Sleep Health and Lifestyle and Sleep Efficiency datasets were used for forecasting sleep disorder into (Insomnia, Sleep Apnea, None) and sleep efficiency respectively. For sleep disorder classification, eight models were evaluated, with XGBoost and LightGBM achieving the highest accuracy of $\mathbf{9 2. 8 6 \%}$ after hyper parameter tuning. In predicting sleep efficiency, five regression models were evaluated where Random Forest outperformed other regression models, achieving a test RMSE of 0.0478 and an $\mathbf{R}^{\mathbf{2}}$ of 0.8772. Future work should focus on integrating these models into wearable devices and healthcare systems to enhance accessibility and usability for end-users.