Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions

2025-06-30

AI Summary

This study addresses the weak modeling of behavioral signals from wearable devices by proposing the first foundation model framework specifically designed for behavioral data. Leveraging real-world wearable data from 162,000 individuals totaling 2.5 billion hours, we introduce a behavior-aware tokenization strategy and a time-scale-adaptive architecture that enables complementary enhancement between behavioral sequences and raw sensor representations. Through self-supervised pretraining followed by multi-task fine-tuning, our model achieves significant performance gains across 57 health prediction tasks—particularly excelling in behavior-driven tasks such as sleep staging. Our core contributions are threefold: (1) the first systematic construction of a behavior-oriented foundation model; (2) the formal establishment of behavioral signals’ intrinsic value and synergistic gain mechanisms in health prediction; and (3) a novel paradigm for dynamic health monitoring grounded in principled behavioral representation learning.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior
• Machine Learning: Deep Neural Architectures and Foundation Models

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

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 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, systematically optimizing architectures and tokenization strategies for this unique dataset. Evaluated on 57 health-related tasks, our model shows strong performance across diverse real-world applications including individual-level classification and time-varying health state prediction. 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.

Problem

• Develop foundation models for behavioral data from wearables
• Optimize architectures for large-scale wearable datasets
• Improve health predictions using behavioral and sensor data

Innovation

• Foundation models for behavioral wearable data
• Optimized architectures and tokenization strategies
• Combined behavioral and raw sensor data representations