A Dynamic Bayesian Network–Based Method for Real-Time Modeling of Communication Behaviors in Medical Internet of Things
Yu Zhang, Zhiyong Hu, Jianjun Xue, Keng Li, Ming Xue, Jingjing Ren, Hao Wu, Jian Zhou, Jiayuan Ling
The rapid proliferation of the Medical Internet of Things (MIoT) has significantly enhanced real-time healthcare monitoring while introducing complex, dynamic communication behaviors among heterogeneous medical devices. Accurate modeling of these behaviors is essential for ensuring network reliability, anomaly detection, and adaptive resource management. However, existing approaches often rely on static statistical models or data-driven techniques that inadequately capture temporal dependencies and probabilistic state transitions in evolving MIoT environments. To address this issue, this paper proposes a Dynamic Bayesian Network (DBN)–based method for real-time modeling of MIoT communication behaviors. The proposed framework represents device states, traffic features, and contextual factors as temporally interconnected nodes within a probabilistic graphical structure, enabling dynamic inference of communication patterns across time slices. An online parameter learning mechanism is incorporated to update transition probabilities under non-stationary network conditions. Experimental evaluations on a representative MIoT communication dataset demonstrate that the proposed DBN-based approach achieves superior behavior prediction accuracy and improved robustness compared with Hidden Markov Models and conventional machine learning baselines, particularly under varying traffic loads. These results indicate that the proposed method provides an effective, interpretable, and scalable solution for real-time communication behavior modeling in MIoT systems.