Anomaly Detection in Adolescent Physical Activity Patterns to Assess Health and Behavioral Risks

Haoran Nie^*

• College of Computer Science and Technology, Shenyang University of Chemical Technology, Shenyang, China

This study proposes a transformative framework for adolescent health assessment by detecting anomalies in physical activity patterns and linking them to potential health and behavioral risks. Central to the methodology is the Adolescent Health Integrator (AHI), a comprehensive model that fuses static attributes (demographic and biological factors), dynamic elements (behavioral trends), risk exposures, and social connections into a unified representation. Leveraging multimodal encoders, graph convolutional networks, and recurrent refinement mechanisms, AHI enables robust inference of health trajectories across diverse adolescent populations while maintaining interpretability and adaptability. Building upon AHI, the Adolescent Health Enhancement Protocol (AHEP) operationalizes these insights into adaptive, context-aware interventions. AHEP incorporates iterative prediction refinement and intervention optimization, dynamically adjusting model outputs and intervention strategies to align with evolving adolescent needs. Its network-aware social propagation mechanism ensures that health improvements spread through peer and community networks, while a personalized prioritization function tailors interventions to individual risk profiles. Experimental validation demonstrates that this integrated approach outperforms traditional models in identifying high-risk behaviors and predicting adverse health outcomes. By combining AHI's comprehensive modeling capacity with AHEP's adaptive intervention strategy, the framework transcends static risk assessment, offering a dynamic and socially responsive pathway for promoting adolescent well-being. This study underscores the potential of combining anomaly detection, graph-based modeling, and personalized health protocols to inform preventive strategies and support data-driven policy decisions in adolescent health research.