Traffic Prediction for Optical Fronthaul Network Using Self-Attention Mechanism-Based Transformer

The research on traffic patterns in advance is significant to improve the transmission performance of delay sensitive services by shortening the delay of queuing and computing in optical networks. However, it is challenging to accurately and rapidly predict changes of traffic in optical networks because of dynamicity and uncertainty of traffic patterns. This paper proposes a model for traffic analysis based on Transformer which can make sure of the accuracy of prediction as fast as possible. Based on the practical traffic data collected from optical fronthaul network, the effectiveness of the model is verified, the prediction error is 5.384%.