LAUSR

Abstract Mitigating globally emerging health problems such as obesity needs scalable solutions that can facilitate health management and promote healthier lifestyles outside of clinical settings. Such scalable solutions, while targeting general population, need to provide personalized behavior change plans that not only fit users’ own underlying physiologic dynamics but also suit their preferences and needs. There has been fast-growing development of mobile health devices and applications for monitoring of human behavior (such as physical activity and food intake) and health status such as BMI. However, there are challenges to translate these noisy and dynamic behavioral data into personalized longitudinal planning. To address such challenges, we develop an integrated framework that unifies dynamic modeling, sparse learning, dictionary learning and matrix completion to translate users’ behavioral data into personalized dynamic system models and use them as constraints for deriving deeply personalized longitudinal health plans. We evaluate the proposed framework on a real-world user behavioral dataset and demonstrate its promising utility and efficacy.