BER Degradation Prediction Using Random Forest Model in GANA Knowledge Plane Platform for 5G/5G-A Transport Network QoS Assurance

• Published in: IEEE/IFIP Network Operations and Management Symposium pp. 1 - 6
• Main Authors: Choi, Taesang, Scheel, Cristian Zumelzu, Park, Moonkook, Kim, Jeongyoon, Chaparadza, Ranganai, Yoon, Sangsik
• Format: Conference Proceeding
• Language: English
• Published: IEEE 12.05.2025
• Subjects: 5G mobile communication; 5G Multi-Layer Transport Network; Agile Management; Autonomic Management & Control (AMC); Degradation; ETSI GANA Knowledge Plane (KP) for AMC for 5G Multi-Layer Transport Network; Knowledge engineering; Maintenance; Prediction algorithms; Predictive models; Quality of service; Software defined networking; Software-Defined Network (SDN); Telecommunication network reliability; Telecommunications
• ISSN: 2374-9709
• DOI: 10.1109/NOMS57970.2025.11073649

Leveraging Random Forest models, we trained our algorithm on real-world data from optical connections and employed a sliding window approach to forecast degradation steps ahead of time. To address the data imbalance inherent in such networks, we applied data cleaning and augmentation techniques. Our model, evaluated on multiple datasets, demonstrated high accuracy in predicting early-stage BER degradation, facilitating proactive maintenance and improving network reliability. The results underline the efficacy of Random Forest models in anticipating service degradation and minimizing disruption in telecommunications infrastructure. The BER Degradation Prediction Using Random Forest Model, together with other AI/ML algorithms required for Autonomic Management & Control (AMC) of a Multi-Layer Transport Network for 5G and Beyond play roles in the AMC operations of an ETSI GANA (Generic Autonomic Networking Architecture) Knowledge Plane (KP) for a Multi-Layer Transport Network. In this paper, we propose GANA Knowledge Plane autonomic management and control implementation architecture and solution for 5G and beyond transport network, specifically focusing on BER degradation prediction based on our ML algorithm.