AI-Driven Network Management: A Review of Machine Learning and Deep Learning Approaches

The AI-based network management has been recognized as an important facilitator for dealing with the increasing complexity, size, and dynamics of modern communication networks. The growth of data communication, heterogeneous networked devices, and stringent network services requires effective network management techniques to ensure the performance, reliability, and security of communication networks. Machine Learning (ML) and Deep Learning (DL) techniques are recognized as intelligent tools to support network management functions such as traffic prediction, resource management, anomaly detection, fault diagnosis, routing control, and security services. This review article discusses the state-of-the-art ML and DL techniques applied to network management, including the methodologies and paradigms of these techniques. The major developments in supervised, unsupervised, and reinforcement learning are discussed to support intelligent decision-making for network management. The review article also addresses the major research challenges of network management using ML and DL techniques. The research challenges are identified as the scarcity of network data, interpretability of ML models,