Experimentation of online models for scalability of 5G network functions

Abrahão Ferreira, Kauan Miranda Tavares, Douglas Almeida Vidal, Silvia Lins, Aldebaro Klautau, Cristiano Bonato Both, Glauco Estácio Gonçalves

DOI: 10.14209/sbrt.2025.1571157085

Evento: XLIII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2025)

Keywords: 5G Core Online Learning Proactive scaling

Abstract

The increasing evolution of 5G networks and the growing demands of new devices, poses significant challenges to the proactive scalability of network functions. Deployed on a non-stationary environment, performance of conventional predictive models drops as new concepts appears, making it necessary to use techniques as online learning, which are still underexplored. This work evaluates and compares eighteen online learning strategies for the scalability of the AMF function. Using real data in scenarios with concept drift, this paper contributes to identifying the more effective approaches. Results indicate that model efficacy is contingent on both predictive accuracy and adaptation latency.

Download