Predicting Sleep Disorders Using a Stacking Ensemble of Gradient Boosting, Decision Tree, and Logistic Regression Models

Abstract

sleep disorders, including insomnia, obstructive sleep apnea, and restless leg syndrome, have significant impacts on human health and productivity. Conventional diagnostic methods, such as polysomnography, are resource-intensive and not suitable for large-scale screening. Recent advancements in machine learning (ML) offer promising alternatives by enabling automated, datadriven diagnosis from lifestyle and health-related parameters. This study presents a comprehensive investigation of existing ML-based approaches for sleep disorder detection, followed by the development of an optimized diagnostic framework. The proposed model integrates advanced preprocessing—feature selection, class balancing via SMOTE-ENN, and normalization—with a Stacking Classifier combining Gradient Boosting, Decision Tree, and Logistic Regression to enhance predictive performance. Evaluation on the Sleep Health and Lifestyle dataset achieved an accuracy of 98.67%, precision of 98.80%, recall of 98.60%, and F1-score of 98.70%, outperforming baseline classifiers and previous literature benchmarks. The framework fills research gaps by making models clearer and easier. It also helps reduce the effect of class imbalance in the data. The design keeps the tool low cost and very easy to use. The system works well with larger data, so it can grow fast. It also helps with the early detection of health problems. The method can support preventive care and also improve patient safety. In this way, the framework is useful for both experts and normal users. It gives support for real healthcare needs and makes solutions simpler.