A Survey of Machine Learning Approach in Fractional Frequency Reuse in 5G Network

Abstract

Fractional Frequency Reuse (FFR) has emerged as a key interference mitigation technique for improving throughput and spectral efficiency in fifth-generation (5G) heterogeneous networks (HetNet), particularly in dense femtocell deployments. However, conventional FFR schemes rely on static or heuristic-based configurations that struggle to adapt to dynamic traffic patterns and complex interference conditions. Recently, machine learning (ML) techniques have been increasingly explored to enhance the adaptability and intelligence of FFR mechanisms. This paper presents a comprehensive survey of machine learning based approaches applied to FFR in 5G networks, with a focus on interference mitigation, throughput enhancement, and Quality of Service (QoS) improvement for cell-edge and indoor users. The study systematically reviews existing ML-driven FFR frameworks, categorizing them based on learning paradigms, network architectures, performance metrics, and deployment scenarios. Key findings reveal that reinforcement learning and deep learning methods are the most widely adopted due to their ability to operate under uncertain and time-varying network conditions. Furthermore, this survey identifies critical research gaps, including scalability challenges, limited real-time implementations, and insufficient consideration of heterogeneous traffic demands. By highlighting recent advances and open research directions, this work provides valuable insights for the development of intelligent, adaptive frequency reuse strategies for future 5G and beyond wireless networks.