Improving Sleep Disorder Diagnosis Through Optimized M

Abstract

Due to their impact on health, sleeping disorders such as obstructive sleep apnea and insomnia should be properly diagnosed. In this work, machine learning models are optimized with the help of the Sleep Health and Lifestyle dataset of Kaggle, which is selected to predict sleep disorders. ANOVA is used in feature selection, and SMOTEENN resampling is used in dealing with imbalance between classes. Additional features are engineered by seven foundation classifiers, including RF, Gradient Boosting, Gaussian Naive Bayes, KNN, DT, LR, and SVM. A number of ML algorithms are then tested. The engineered Voting Classifier is the most successful, as it has accuracy. Accuracy of Stacking Classifier is lower in the event of using original features. Making decisions about the model using Explainable AI tools such as LIME and SHAP enhances the transparency and makes clinical decisions.