Exploring Sleep Disorders: A Comparative Analysis of Machine Learning Algorithms on Sleep Health and Lifestyle Data

Classifying sleep disorders is crucial for improving human health, as conditions like apnea significantly impact well-being. Traditional methods are often error-prone, necessitating accurate machine learning algorithms (MLAs). This research compares several MLAs using the Sleep Health and Lifestyle Dataset. The dataset is processed by label encoding variables such as gender, occupation, blood pressure, and BMI category, followed by MinMax scaling for normalization. It is splitted into training, testing, and validation sets. K-fold cross-validation is utilized for reliable performance estimation, and hyperparameter tuning is conducted to enhance model accuracy. The parameters tuned include the number of estimators, learning rate, and maximum depth for algorithms like Random Forest, AdaBoost, and Gradient Boosting. Our findings indicate that Gradient Boosting achieved the highest classification accuracy at 93.80%, outperforming Random Forest (90.26%), Logistic Regression (91.15%), and AdaBoost (91.15%). Gradient Boosting's superior performance is attributed to its ability to sequentially build and optimize weak learners, effectively capturing complex patterns in the data. This study underscores the effectiveness of optimized MLAs in accurately classifying sleep disorders, emphasizing the importance of thorough preprocessing and hyperparameter tuning.