A survey of the integration between machine learning and artificial intelligence techniques in software-defined...

Recently, Software-Defined Networking has become one of the key technologies in modern telecommunication networks. Its programmable and flexible architecture enables centralized control of network resources, automation of management processes, and integration with other technologies such as Network Function Virtualization, Edge/Fog/Cloud Computing, and 5G/6G networks. Moreover, SDN provides a natural environment for implementing advanced Machine Learning and Artificial Intelligence techniques, which are increasingly applied to anomaly detection, network traffic analysis, load prediction, and flow optimization. This article presents a comprehensive survey of the integration between SDN and AI across diverse networking domains, including core, cloud, edge, wireless, IoT, and vehicular networks. More than 400 publications from 2015 to 2025 have been analyzed and classified according to network domain, application area, and category of AI techniques employed-particularly supervised, unsupervised, and hybrid learning methods. Based on this systematic analysis, a mapping framework was developed to illustrate the relationships between AI techniques and SDN application domains, revealing dominant research trends and existing gaps in the current literature. Unlike many previous surveys that focus on individual SDN aspects, selected network domains, or isolated AI-based applications, this survey provides an integrated cross-domain perspective on AI-enabled SDN. Its main contribution lies in a taxonomy-driven mapping of machine learning paradigms, SDN application areas, and network domains, which clarifies how current research supports the transition from programmable SDN toward more autonomous and intelligent network management. The results of the conducted review indicate that the integration of SDN and AI constitutes a foundation for the development of intelligent network management mechanisms. The machine learning techniques presented in this paper form the basis of the currently emerging AI-driven solutions for SDN environments, supporting prediction, classification, anomaly detection, and network performance optimization. At the same time, it should be emphasized that the supervised, unsupervised, and hybrid paradigms primarily address prediction, classification, detection, and adaptive optimization functions, whereas full network automation also requires interactive learning mechanisms, continuous decision-making, multi-domain orchestration, and closed-loop control. In the final part of the article, the main directions for future research related to the development of more autonomous, adaptive, and explainable network management architectures are also identified.