Learning to Classify DWDM Optical Channels from Tiny and Imbalanced Data.

Paweł Cichosz,Stanisław Kozdrowski,Sławomir Sujecki

doi:10.3390/e23111504

Abstract

Applying machine learning algorithms for assessing the transmission quality in optical networks is associated with substantial challenges. Datasets that could provide training instances tend to be small and heavily imbalanced. This requires applying imbalanced compensation techniques when using binary classification algorithms, but it also makes one-class classification, learning only from instances of the majority class, a noteworthy alternative. This work examines the utility of both these approaches using a real dataset from a Dense Wavelength Division Multiplexing network operator, gathered through the network control plane. The dataset is indeed of a very small size and contains very few examples of “bad” paths that do not deliver the required level of transmission quality. Two binary classification algorithms, random forest and extreme gradient boosting, are used in combination with two imbalance handling methods, instance weighting and synthetic minority class instance generation. Their predictive performance is compared with that of four one-class classification algorithms: One-class SVM, one-class naive Bayes classifier, isolation forest, and maximum entropy modeling. The one-class approach turns out to be clearly superior, particularly with respect to the level of classification precision, making it possible to obtain more practically useful models.

Highlights

Growing traffic in backbone networks makes dynamic and programmable optical networks increasingly important
R package [53], with a custom prediction method to handle the one-class classification mode implemented for this work
The work has provided additional evidence that applying machine learning to optical channel classification is a promising work direction, but is associated with important challenges

Summary

Introduction

Growing traffic in backbone networks makes dynamic and programmable optical networks increasingly important. This applies to Dense Wavelength Division Multiplexing (DWDM) networks whereby efficient use of network resources is of paramount importance. In modern DWDM optical networks, following the software defined network paradigm, DWDM network reconfiguration is becoming more frequent, making the evolving network more resilient and adapting faster to real changes in bandwidth demand so that network reconfigurations can closely match changes in bandwidth demand. Bandwidth demand can change very quickly (fluctuations can occur within minutes), while network reconfigurations typically take much longer. This is mainly due to operational processes that are too slow to allow real-time network re-optimization. It is important that DWDM network reconfigurations are automated and as fast as possible, without significantly increasing operational costs

Objectives

Methods

Results

Discussion

Conclusion