Machine Learning-Driven Stress Prediction: A Comparative Analysis and Web Application Using the Sleep Health and Lifestyle Dataset

Stress significantly affects both physical and mental health, making early detection crucial for preventing severe complications. This study applies machine learning models to the Sleep Health and Lifestyle dataset to predict stress levels using physiological and behavioral indicators. Extensive data preprocessing, including feature selection and normalization, has been performed to ensure model accuracy. The dataset was split into training and testing subsets, and five machine learning models (i.e., Random Forest, Support Vector Machine (SVM), XGBoost, Decision Tree, and Logistic Regression) were evaluated. The results highlight variations in predictive performance, with some models achieving high accuracy in stress classification. Feature correlation analysis identified sleep duration, physical activity, and heart rate as key stress predictors. Finally, this study proposes a web application that enables real-time stress prediction by allowing users to input personal data and receive instant feedback. This application enhances personalized health monitoring and supports AI-driven stress detection, contributing to mental health awareness and early intervention strategies.