A Comprehensive Review of Prediction of IoT Traffic Using Gradient Boosting, Auto-Metric Graph Neural Network, and Lyapunov Optimization-Based Predictive Model

Abstract

The exponential growth of Internet of Things (IoT) devices has led to a massive surge in heterogeneous network traffic, posing significant challenges for efficient traffic prediction and resource management. Accurate IoT traffic prediction is essential for optimizing network performance, reducing latency, and ensuring Quality of Service (QoS) in dynamic environments. Traditional statistical and rule-based methods fail to capture the complex spatio-temporal dependencies and non-linear characteristics inherent in IoT networks. Consequently, advanced machine learning and optimization techniques such as Gradient Boosting, Graph Neural Networks (GNNs), and Lyapunov optimization have emerged as promising solutions. This paper presents a comprehensive review of state-of-the-art approaches for IoT traffic prediction, focusing on hybrid models that integrate Gradient Boosting algorithms, Auto-Metric Graph Neural Networks, and Lyapunov-based optimization frameworks. Gradient Boosting techniques enhance predictive accuracy by leveraging ensemble learning, while GNNs effectively model spatial dependencies in network topologies. Lyapunov optimization provides a mathematical framework for dynamic resource allocation and system stability in stochastic environments. The study analyses recent advancements in IoT traffic prediction, highlighting model architectures, datasets, evaluation metrics, and performance improvements. A comparative analysis is conducted to evaluate the strengths and limitations of different approaches. Furthermore, challenges such as scalability, real-time processing, and data privacy are discussed. The review concludes with future research directions emphasizing hybrid deep learning models, federated learning, and intelligent edge computing for next-generation IoT systems.