Evaluating Cross- feature Trained Machine Learning Models for Estimating QoT of Unestablished Lightpaths

Abstract

The rapid increase in bandwidth-driven applications has resulted in exponential internet traffic growth, especially in the backbone networks. To address this growth of internet traffic, operators always demand the total capacity utilization of underlying infrastructure. In this perspective, precise estimation of the quality of transmission (QoT) of the lightpaths (LPs) is vital for reducing the margins provisioned by uncertainty in network equipment's working point. This article proposes and compares several data-driven Machine learning (ML) based models to estimate QoT of unestablished LP before its deployment in the future deploying network. The proposed models are cross-trained on the data acquired from an already established LP of an entirely different in-service network. The metric considered to evaluate the QoT of LP is the Generalized Signal-to-Noise Ratio (GSNR). The dataset is generated synthetically using well tested GNPy simulation tool. Promising results are achieved to reduce the GSNR uncertainty and, consequently, the provisioning margin.