An ML-driven framework for continuous wearable health surveillance and energy-efficient communication

Existing wireless body area network (WBAN) communication schemes often depend on fixed event assumptions, which results in redundant transmissions, poor energy utilization, and limited scalability in continuous health monitoring systems. To overcome these limitations, this paper introduces an machine learning-assisted (ML-assisted) Bitmap-Assisted MAC framework that employs data-driven transmission probability estimation to enable selective and energy-efficient communication. Machine learning models are trained in a MATLAB-based environment using wearable health data to predict transmission relevance, and the estimated probabilities are integrated into CC2420 ZigBee energy model for MAC-layer analysis. The proposed approach improves channel utilization by activating only transmission-relevant BAN devices while suppressing non-critical traffic. Simulation results demonstrate that the proposed method achieves 35-60% energy reduction compared to conventional MAC schemes, with the Gradient Boosting-assisted variant providing the best performance in terms of probability estimation accuracy and overall energy efficiency.