A Machine Learning Framework for Predicting Sleep Disorders Using Lifestyle and Health Factors
Students & Supervisors
Student Authors
Supervisors
Abstract
Approximately one-third of the population worldwide is affected by sleep disor- ders. Several machine-learning (ML) studies on sleep disorders are binary and based on small datasets with little clinical interpretability. This research discusses a multi-class ML framework, utilizing demographic, lifestyle, and physiological data. To develop this framework, it applied clinically informed feature engineering such as Mean Arterial Pressure (MAP), Hypertension Stage, and a novel Cardio-Metabolic Stress Index (CSI) in conjunction with SMOTE and feature selection using RFECV, and hyperparameter tuning via Optuna. The Gradient Boosting Classifier exhibited the most robust performance. To explain feature importance, SHAP analyses iden- tified sleep duration, heart rate, BMI, daily steps, and the Cardio-Metabolic Stress Index (CSI) as the strongest predictive features of sleep disorders. Future expansion will include validation with real-world datasets in clinical populations to increase the generalizability of the model for deployment in clinical and healthcare systems.
Keywords
Publication Details
- Type of Publication:
- Conference Name: International Conference on Intelligent Data Analysis and Applications (IDAA 2025)
- Date of Conference: 12/12/2025 - 12/12/2025
- Venue: Daffodil International University
- Organizer: Department of Computer Science and Engineering (CSE), Daffodil International University (DIU)