Foundation Models for Wearable Behavioral Data Improve Health Predictions
By
brandonb
9mo ago· 2 min readenInsight
75/100
Toasty
Bagelometer↗
Lightly browned and well buttered. A solid pick from the rack.
Score75TypeanalysisSentimentpositive
Summary
Researchers developed foundation models for behavioral data from wearable devices using over 2.5 billion hours of data from 162,000 individuals. The models focus on behavioral signals rather than raw sensor data, as behavioral data aligns better with physiologically relevant timescales. Evaluated on 57 health-related tasks, the model showed strong performance across diverse applications including sleep prediction and time-varying health state prediction, with further improvements when combined with raw sensor data representations.
Key quotes
· 4 pulledBehavioral data often being more informative due to their alignment with physiologically relevant timescales and quantities
We develop foundation models of such behavioral signals using over 2.5B hours of wearable data from 162K individuals
The model excels in behavior-driven tasks like sleep prediction, and improves further when combined with representations of raw sensor data
These results underscore the importance of tailoring foundation model design to wearables and demonstrate the potential to enable new health applications
Wearable devices record physiological and behavioral signals that can improve health predictions. While foundation models are increasingly used for such predictions, they have been primarily applied to low-level sensor data, despite behavioral data often
