![[2602.18521] AdaptStress: Online Adaptive Learning for Interpretable and Personalized Stress Prediction Using Multivariate and Sparse Physiological Signals](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.18521] AdaptStress: Online Adaptive Learning for Interpretable and Personalized Stress Prediction Using Multivariate and Sparse Physiological Signals
Summary
The paper presents AdaptStress, a novel model for predicting stress levels using physiological data from wearables, achieving superior accuracy over traditional methods.
Why It Matters
As stress management becomes increasingly important for mental health, this research highlights the potential of consumer-grade wearables in providing personalized, real-time stress predictions. By leveraging advanced machine learning techniques, it paves the way for scalable mental health monitoring solutions.
Key Takeaways
- AdaptStress model outperforms traditional forecasting methods in stress prediction.
- Utilizes multivariate physiological signals from consumer wearables for personalized insights.
- Demonstrates the importance of sleep metrics as key predictors of stress.
- Individual-specific patterns reveal varying effects of identical features across users.
- Supports the feasibility of continuous mental health monitoring in real-world settings.
Computer Science > Machine Learning arXiv:2602.18521 (cs) [Submitted on 19 Feb 2026] Title:AdaptStress: Online Adaptive Learning for Interpretable and Personalized Stress Prediction Using Multivariate and Sparse Physiological Signals Authors:Xueyi Wang, Claudine J. C. Lamoth, Elisabeth Wilhelm View a PDF of the paper titled AdaptStress: Online Adaptive Learning for Interpretable and Personalized Stress Prediction Using Multivariate and Sparse Physiological Signals, by Xueyi Wang and 2 other authors View PDF HTML (experimental) Abstract:Continuous stress forecasting could potentially contribute to lifestyle interventions. This paper presents a novel, explainable, and individualized approach for stress prediction using physiological data from consumer-grade smartwatches. We develop a time series forecasting model that leverages multivariate features, including heart rate variability, activity patterns, and sleep metrics, to predict stress levels across 16 temporal horizons (History window: 3, 5, 7, 9 days; forecasting window: 1, 3, 5, 7 days). Our evaluation involves 16 participants monitored for 10-15 weeks. We evaluate our approach across 16 participants, comparing against state-of-the-art time series models (Informer, TimesNet, PatchTST) and traditional baselines (CNN, LSTM, CNN-LSTM) across multiple temporal horizons. Our model achieved performance with an MSE of 0.053, MAE of 0.190, and RMSE of 0.226 in optimal settings (5-day input, 1-day prediction). A comparison with...
