AI-Driven Network Traffic Prediction and Optimization: A Systematic Review of Machine Learning and Deep Learning Approaches

Abstract

Efficient network traffic management is essential for maintaining high performance, reliability, and quality of service in modern communication infrastructures. The increasing volume and variability of network traffic make conventional statistical and rule-based prediction methods insufficient for many contemporary networking environments. Machine Learning (ML) and Deep Learning (DL) have therefore gained considerable attention for predicting traffic patterns and optimizing network resources. This review systematically examines AI-based approaches for network traffic prediction, including regression models, decision trees, support vector machines, recurrent neural networks, long short-term memory networks, convolutional neural networks, transformers, and hybrid architectures. Existing studies are analyzed with respect to traffic characteristics, datasets, prediction horizons, evaluation metrics, and deployment environments. The review also investigates how traffic prediction can support routing optimization, congestion control, bandwidth allocation, load balancing, and quality-of-service management. Particular attention is given to the challenges of concept drift, highly dynamic traffic, data imbalance, computational complexity, and model interpretability. The paper concludes by identifying opportunities for transformer-based models, federated learning, and real-time AI-enabled traffic engineering.