Accurate Human Stress Detection from Sleeping Habits Using a Hybrid Ensemble of Machine Learning Models

Preventing a variety of health problems and preserving well-being depend on effective stress identification and management. Recently, machine learning has gained attention for its potential in stress detection, especially by analyzing physiological and behavioral indicators. This paper presents a new framework for human stress detection by examining sleeping habits through an advanced hybrid machine learning (ML) model. The proposed model integrates five diverse algorithms—Logistic Regression (LR), Gradient Boosting (GB), Random Forest (RF), AdaBoost (AB), and Support Vector Classifier (SVC)—using a soft voting mechanism. By combining these models, the approach capitalizes on each algorithm's strengths, enabling it to detect subtle patterns in sleep metrics that may indicate stress. Key features used in the model include sleep duration, snoring rate, respiration rate, and body temperature. We tested the model on a dataset containing sleep information from individuals with different stress levels. Results show that this hybrid approach surpasses individual models in terms of accuracy, recall, precision, and F1 score, establishing a reliable tool for stress detection based on sleep behavior.