Application of ML Algorithms for Prediction of the QoT in Optical Networks with Imbalanced and Incomplete Data

Telecommunication optical networks grow steadily in response to increasing traffic. Thus, to accommodate a steady traffic growth, network operators need to increase the number of active optical channels and reconfigure networks constantly. These tasks cannot be successfully performed without an assessment of the quality of transmission prior to any changes made to an optical network. In this contribution a machine learning-based method for the assessment of the quality of transmission in a Dense Wavelength Division Multiplexing (DWDM) network is proposed. The proposed machine learning method uses a database, which was created only on the basis of information that is available to a network operator via the DWDM network control plane. Particular focus of this study is given to one-class classification algorithms: the one-class support vector machine, one-class naive Bayes classifier and isolation forest algorithms. This is to spare scarcely available examples of unsuccessful channel configurations for the purpose of model evaluation and learn exclusively from successful channel examples. The results obtained show that one-class classification algorithms present a viable alternative to the more common binary classification approach.