AI-DRIVEN PREDICTIVE HEALTH INTELLIGENCE FOR SMART CITIES: MODELING URBAN STRESS AND HEALTH RISKS USING POI AND MOBILITY DATA

Abstract

Urbanization continues to exacerbate health disparities in rapidly growing cities, particularly among underserved populations. This study introduces a predictive health intelligence framework that integrates Points of Interest (POI), mobility patterns, and socioeconomic indicators to model urban stress and health risks. The proposed system includes the development of the Urban Stress Index (USI) , a composite geospatial metric and HealthPOI-Net (Predictive Optimization Interface for Health Data), an AI-driven tool designed to visualize and forecast risk zones. Using Dubai as a pilot site due to its rapid growth, high digital infrastructure, and demographic diversity, the model demonstrated 86.3% accuracy in forecasting health risk categories (high/medium/low) when compared against anonymized emergency health service data and hospitalization trends. The system also showed significant alignment with aggregated hotline engagement trends and employer-submitted wellness indicators, reinforcing the validity of the stress-risk predictions. This framework offers near real-time insights and simulation capabilities for urban planners, public health officials, and policymakers, enabling data-informed interventions tailored to vulnerable populations. While not yet globally deployed, the model presents a scalable and ethically grounded prototype that may support future smart city health strategies, particularly in the Global South. The study emphasizes ethical AI design, including compliance with GDPR and privacy-by-design principles for all geospatial and behavioral data, with attention to data opt-in mechanisms, transparency, and the inclusion of marginalized populations in model validation and interpretation.