Predicting the Next-Day Perceived and Physiological Stress of Pregnant Women Using Machine Learning and Explainability

Summary: We build explainable ML models to predict next-day perceived and physiological stress for pregnant individuals.

Abstract

We extract daily features from wearables and ecological momentary assessments to forecast next-day stress. Models achieve consistent predictive utility and are accompanied by post-hoc explainability (e.g., SHAP) to highlight actionable drivers such as sleep and activity patterns. The analysis underscores interpretable pathways for digital monitoring and personalized care during pregnancy.