Temporal Dynamics of Classroom Stress: Insights from Wearable Sensors and Machine Learning

Stress among college students significantly affects academic performance and well-being, yet traditional detection methods often fall short. This study explores the potential of wearable sensors-specifically Hexoskin vests and Actigraph watches-for predicting stress levels among college students. Over 18 weeks, biometric data, including heart rate, respiration rate, breathing volume, and movement, were collected from five undergraduates during classroom sessions. We applied classification algorithms such as Random Forest, K-Nearest Neighbors, XGBoost, and Neural Networks to analyze the temporal dynamics of stress. The Random Forest model achieved the highest accuracy, with a median F1 score of 92 %, using a 30-day sliding window for stress classification. Breathing volume emerged as the most significant predictor, followed by heart rate, respiration rate, and movement. Individualized models consistently outperformed generalized ones, highlighting the variability in students' stress responses. This study underscores the potential of biometric data from wearable sensors for real-time stress prediction in academic settings, while advocating for further research with larger and more diverse populations to validate and expand these findings.