Deep Learning-enabled Multi-step Spatio-temporal IP Traffic Streams Prediction in IP-over-WDM-based Backbone Optical Networks

We propose a deep learning (DL) enabled multi-step spatio-temporal IP traffic streams prediction framework in IP-over-WDM-based backbone optical networks. The experimental results show that it greatly improves the prediction accuracy of inter-city IP traffic streams in a real network compared with baseline DL models.