Supervised Machine Learning Techniques for Quality of Transmission Assessment in Optical Networks

Abstract

We propose and compare a number of machine learning models to classify unestablished lightpaths into high or low quality of transmission (QoT) categories in impairment-aware wavelength-routed optical networks. The performance of these models is evaluated in long haul communication networks and compared to previous proposals. Results show that, especially random forests and bagging trees approaches, significantly reduce the required computing time to classify the QoT of a given lightpath, while accuracy remains around 99.9%.