p09 / Journal
Federated learning for 5G base station traffic forecasting
Computer Networks, 235, 109950 · 2023
In plain language
Can base stations forecast network demand together while keeping their traffic measurements local? This study tests federated learning on real LTE observations.
Research context
It connects distributed learning with practical network planning and evaluates its computational and environmental costs.
Approach
Five neural architectures, non-identically distributed base-station data, preprocessing, aggregation comparisons, and a larger synthetic scenario.
Findings
Federated models achieve prediction errors comparable to centralized training. Preprocessing helps, while more elaborate aggregators do not outperform simpler ones in these experiments.
Summary based on abstract.