Recent Advances in Prediction of IoT Traffic Using Gradient Boosting, Auto-Metric Graph Neural Network, and Lyapunov Optimization-Based Predictive Model: A Systematic Review

Abstract

The rapid growth of the Internet of Things (IoT) has led to an unprecedented increase in network traffic, making efficient traffic prediction a critical requirement for ensuring Quality of Service (QoS), resource optimization, and network stability. Traditional statistical models are often inadequate for capturing the dynamic, nonlinear, and heterogeneous nature of IoT traffic. Consequently, advanced machine learning and optimization-based techniques such as Gradient Boosting, Graph Neural Networks (GNNs), and Lyapunov Optimization have gained significant attention in recent years. This systematic review presents a comprehensive analysis of recent advancements in IoT traffic prediction, focusing on hybrid and intelligent models that integrate learning-based and optimization-driven approaches. Gradient boosting techniques provide robust predictive performance by handling nonlinearity and feature interactions, while Auto-Metric Graph Neural Networks effectively capture spatial-temporal dependencies inherent in IoT networks. Furthermore, Lyapunov optimization-based predictive models enable dynamic resource allocation and system stability by transforming optimization problems into queue stability formulations.  The review analyses recent studies (2020–2023), highlighting their methodologies, datasets, evaluation metrics, and performance improvements. It also identifies key challenges such as scalability, data heterogeneity, real-time adaptability, and energy efficiency. Additionally, the integration of deep learning with optimization frameworks is explored as a promising direction for next-generation IoT systems. The findings suggest that hybrid models combining machine learning with optimization techniques outperform traditional approaches in terms of prediction accuracy, latency reduction, and network efficiency. This review provides insights into emerging trends, comparative analysis, and future research directions for intelligent IoT traffic prediction systems.