p08 / Conference
Towards Energy-Aware Federated Traffic Prediction for Cellular Networks
Fog and Mobile Edge Computing (FMEC 2023), 93-100 · 2023
In plain language
Accurate traffic prediction has an environmental cost. This paper evaluates that trade-off for federated cellular-network forecasting.
Research context
Model selection should consider energy and emissions alongside forecast accuracy.
Approach
A sustainability indicator and comparisons of deep-learning architectures using real base-station measurements from Barcelona.
Findings
Larger models provide only modest predictive improvements while increasing the estimated carbon footprint substantially.
Summary based on abstract.