Sudden Traffic Prediction Based on Gated Recurrent Unit in Low Earth Orbit Satellite Networks

As an important infrastructure to provide global network services for worldwide users, Low Earth Orbit (LEO) satellite networks with wide coverage are facing uneven traffic patterns due to the diverse user and requirement distribution. Thus, effective traffic prediction (especially sudden traffic prediction) is critical for network operators to optimize network performance. This paper proposes a novel method for sudden traffic prediction based on Gated Recurrent Units (GRU) in LEO Satellite networks. In the proposed method, the in-band network telemetry (INT) technology is first adopted to gather hop latency data which is used to predict the traffic. We then develop a prediction model for sudden traffic based on GRU, including two stages, i.e., offline training and real-time prediction. Finally, the numerical results are conducted to demonstrate the proposed method.