This chapter investigates the pivotal role of Network Slicing (NS) in
optimizing the performance of Fifth-Generation (5G) mobile networks, which offer
unprecedented connectivity through vast bandwidth, ultra-low latency, and scalability
to support diverse Quality of Service (QoS) requirements. NS allows the segmentation
of a physical network into multiple logical networks, each tailored to specific
application needs, from mission-critical industrial automation to high-bandwidth video
streaming. To classify user requests into the most appropriate Network Slices (NSs),
the study applies Machine Learning (ML) models trained on an extensive 5G traffic
dataset. Utilizing classical ML methods and advanced ensemble techniques, such as
Support Vector Classifier (SVC), K-Nearest Neighbors (KNN), XGBoost, and Random
Forest, the models achieve an impressive accuracy of 98.51%, outperforming previous
benchmarks. The research contributes to the field through meticulous data
preprocessing, including feature selection and oversampling to address class imbalance,
as well as rigorous feature engineering to enhance model interpretability and
efficiency. Cross-validation and hyperparameter optimization further ensure model
robustness. The results highlight the critical role of ML-driven predictive modeling in
improving NS deployment, resource allocation, and QoS management in 5G networks,
thus enabling more efficient service delivery and a superior user experience.
Keywords: 5G networks, Cross-validation, Data preprocessing, Feature selection, Hyperparameter optimization, Machine learning, Model evaluation, Network slicing, Predictive modeling, Quality of Service (QoS).
