Comparative clustering strategies for network slicing

As 5G evolves, effective network slice design is crucial for efficient resource allocation and optimal service quality. This study compares eight clustering strategies across architecture, data type, and algorithm using real 5G data. Results show single-layer clustering achieves top metrics, but two-layer clustering improves distribution and fairness. Combining wireless and service-side features enhances balance, while a two-layer model using K-means and HDBSCAN offers optimal stability and interpretability. These insights provide practical guidance for building more robust and equitable network slicing strategies.