A Systematic Review of Spectral Graph Methods for 5G Core Network Slicing: Methods, Architectures, and Future Research Directions

Abstract

The rapid evolution of fifth-generation mobile communication systems has introduced network slicing as a key paradigm for supporting heterogeneous services with diverse quality-of-service requirements. It enables the logical partitioning of shared physical infrastructure into multiple virtual networks tailored for applications such as enhanced mobile broadband, ultra-reliable low-latency communications, and massive machine-type communications. Despite its potential, efficient resource allocation, slice isolation, and dynamic orchestration remain significant challenges due to the scale and complexity of modern core networks. Spectral graph methods have emerged as effective tools for modeling and optimizing network structures by leveraging eigenvalues and graph Laplacians to capture topological properties. This review examines spectral graph-based approaches in network slicing, focusing on techniques such as spectral clustering, graph neural networks, Laplacian-based optimization, and hybrid machine learning frameworks. It also explores their integration with enabling technologies like software-defined networking, network function virtualization, and edge computing. Findings suggest that spectral methods improve scalability, adaptability, and resource efficiency in dynamic environments. However, challenges such as computational overhead and real-time implementation persist, indicating the need for advanced, energy-efficient, and AI-driven orchestration strategies in next-generation networks.