Social Media-Driven Stress and Sleep Analysis Using QPSO-Enhanced Explainable AI Models

Abstract

The growing use of digital technology in everyday life is notably through social networks, which has led to a worldwide surge in stress levels and disturbances of sleep quality. This work introduces a machine learning model for predicting sleep quality and stress based on social media usage patterns, lifestyle factors, and physiological signals. Four regression-based machine learning models are Random Forest, Gradient Boosting, Support Vector Regression, and Linear Regression. They were run on a dataset that includes behavioral (for example, inbed screen use), physiological (e.g., cortisol and melatonin levels), and self-reported health measures. Hyperparameter tuning was performed to improve performance using Quantum- Behaved Particle Swarm Optimization (QPSO), which resulted in substantive improvements in model accuracy and generalizability. In addition, explainable artificial intelligence (XAI) methods such as SHAP and LIME were used to explain the predictions of the models and determine the primary predictors that influence sleep and stress outcomes. The findings show that pre-sleep social media consumption, sleep latency, and stress ratings are good predictors of sleep quality and vice versa. The suggested ML-QPSO-XAI model improves predictive reliability and transparency, thus proving an effective tool for upcoming digital health devices and mental well-being monitoring systems.