With ever-increasing traffic, the need more dynamic, flexible and autonomous optical networks is more important than ever. The availability of performance monitoring data makes it possible to leverage machine learning (ML) for fast quality of transmission (QoT) estimation and performance prediction of lightpaths in complex optical networks. In this work, we explore classifiers based on support vector machine (SVM) and artificial neural network (ANN) for QoT estimation of unestablished lightpaths. Using a synthetic knowledge base (KB), the classification accuracy of the ANN and SVM models decreased from 99%, with a complete feature set, to 85.03% and 88.52%, respectively, with a reduced feature set. We also propose a Long Short-Term Memory (LSTM), an Encoder-Decoder LSTM and a Gated Recurrent Unit (GRU) models, trained with 13-month field performance data, for lightpath signal-to-noise (SNR) prediction over forecast horizons up to 4 days. Positive R 2 values combined with low ( 0.285 dB) root mean square error (RMSE) indicated that the GRU model achieved slightly better predictions than the naive method for forecast horizons ranging from 1 to 96 hours, whereas the LSTM performed better over 24 to 96-hour forecast horizons. The Encoder-Decoder LSTM model achieved the lowest R 2 and the highest RMSE values (0.296 dB). Additional input data will be needed to improve the prediction accuracy of the LSTM and GRU models trained with single lightpath data.

No abstract

We show how the Recurrent Neural Networks can be used for performance prediction of lightpaths using field bit error rate data. Moreover, we illustrate how the forecast horizons and observation windows affect the forecast accuracy.

No abstract

Machine Learning (ML) is emerging as a promising solution for managing the physical layer complexity of heterogeneous dynamic optical networks transporting multiple applications in a software defined network (SDN) context, namely for performance prediction. We propose two multivariate neural network models based on Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) methods, trained with field performance data and features, for predicting lightpath signal-to-noise ratio (SNR) over forecast horizons of up to 4 days. The best performance is achieved by using a 5-feature LSTM multivariate model over forecast horizons of up to 96 hours, with an absolute maximum error (AME) of 0.90 dB, compared to 0.91 dB and 0.97 dB for the GRU and LSTM univariate models, respectively, and 1.21 dB for a persistence model. The 2-feature multivariate models obtained through feature engineering perform better than their univariate counterparts for forecast horizons of up to 40 hours. Lastly, we explore the concept of transfer learning (TL) by testing the trained multivariate LSTM and univariate GRU models on field data from two lightpaths carried on the same route. The TL models underperform the naive model for the lightpath carried in a different optical fiber. However, for the lightpath carried in the same optical fiber on a portion of the same route, the LSTM-based TL model outperforms the naive model with a difference of up to 0.11 dB at a 96-hour forecast horizon, compared to 0.30 dB for the lightpath in the source domain, while using 3 times less training data.