HealthCAT: An Interpretable Encoder-only Transformer Framework for Health Indicator Prediction and Temporal Interpretation of Wearable Sensor Data

2026-07-29

AI Summary

This study addresses the limited interpretability of existing deep learning approaches in wearable sensor data analysis, which prioritize prediction accuracy over explanatory insight and thus hinder behavioral understanding and intervention design in health research. To bridge this gap, the authors propose HealthCAT, a novel framework that tightly integrates timestep-level interpretability with health prediction tasks. By leveraging an encoder-only Transformer architecture enhanced with Attentive Class Activation Tokens (AttentiveCAT), HealthCAT simultaneously achieves high predictive performance and generates class-specific temporal explanations aligned with domain-relevant behavioral cycles, such as time-of-day patterns. Evaluated on two real-world datasets comprising 306 participants, the method demonstrates statistically significant improvements—up to 17% higher F1 score and 12% greater accuracy (p<0.05)—over baseline models. Masking experiments further validate that the identified timesteps carry substantial predictive value.

Abstract

Wearable sensors continuously capture fine-grained multivariate time-series data, providing opportunities to model behavioural patterns associated with health outcomes. However, existing deep learning methods prioritise predictive accuracy over interpretability, limiting their application in health research. In this study, we present HealthCAT, a flexible framework that integrates an Encoder-only Transformer with an Attentive Class Activation Token (AttentiveCAT) to generate class-specific, time-step-level interpretations. These interpretations can be mapped back onto behavioural cycles that are relevant to the domain (e.g., time-of-day), supporting individual-level analysis of wearable sensor data. We evaluated HealthCAT using two real-world wearable sensor datasets (306 participants in total). HealthCAT outperformed deep learning baselines by up to 17\% in F1-score and 12\% in accuracy on both datasets ($p<0.05$). In masking experiments, the time steps identified by HealthCAT carried significantly more predictive value than random selection across all masking conditions ($p<0.05$), indicating that the identified time steps are predictively informative. By coupling predictive performance with validated time-step-level interpretability, HealthCAT moves wearable sensor analysis beyond aggregated metrics towards temporal patterns that support health monitoring, behavioural pattern analysis, and intervention design in health research. The significance of this work is that it enables accurate prediction of health indicators from wearable sensor data while providing insights into when and how physical activity patterns occur, rather than relying solely on aggregated summary measures.

Problem

• interpretability
• wearable sensor data
• health indicator prediction
• time-series analysis
• temporal patterns

Innovation

• Interpretable Transformer
• Attentive Class Activation Token
• Wearable Sensor Data
• Time-step-level Interpretability
• Health Indicator Prediction