Multivariate Stress Forecast from Sparse Data during Lifestyle Interventions

We developed a protocol using a commercial Garmin smartwatch for stress forecasting, validated through a leave-one-subject-out approach, demonstrating robustness and effective use of sparse data from wearable devices. The model excels with a one-day prediction window and a three-day training window, sacrificing a bit of performance over extended predicting windows. This demonstrates the potential of wearable technology for noninvasive, real-time health management. Our findings proved the adaptability of our model to senior populations and its practicality for individuals requiring continuous stress monitoring. This work contributes significantly to the development of accessible, proactive mental health tools and sets a foundation for future research into enhancing the accuracy and responsiveness of stress forecasting.