Query Based Iterative Learning Approach for Lightpath Deployment in Optical Networks

Predicting the Quality of Transmission (QoT) of a Lightpath (LP) before its actual deployment is important for efficient resource utilization. Conventionally, analytical models using closed-loop formulation estimate QoT, which imposes substantial margins to avoid network outages. Recently, data-driven techniques have been shown as a potential alternative with excellent precision and real-time applicability. However, data-driven techniques require sufficient training data, which might be challenging to acquire during real network operations. In this context, we proposed a novel unsupervised Iterative learning (IL) framework developed on top of the Random forest (RF) classifier for QoT estimation of LP before deployment. We considered the Generalized signal-to-noise ratio (GSNR) as a characterizing parameter for QoT estimation of LP. Our simulation results illustrate that, by employing the proposed iterative learning approach, we can obtain 99% classification accuracy with a reduced number of training samples compared to the traditional supervised learning approach.