Search

Chronic stress has significant long-term behavioral and physical health consequences, including an increased risk of cardiovascular disease, cancer, anxiety and depression. This paper conducts post-hoc experiments and simulations to demonstrate feasibility of both real-time stress forecasting and stress intervention adaptation and optimization. Using physiological data collected by ten individuals in the natural environment for one week, we first show that simple Hidden Markov Models can be used to forecast heart rate variability - a proxy for stress - up to 3 minutes in advance. Second, we expand Q-Learning (QL), a reinforcement learning methodology previously used to adapt non-pervasive interventions, to take advantage of the always-available nature of pervasive technology. Using eligibility traces, we demonstrate how QL could be used by a pervasive health system to adapt and deliver any number and type of interventions for a given health event. Our hope is that this work will take us one step closer to using pervasive devices to assist in the daily management of chronic stress and other health challenges.