Importance of Optical Metrics for IGP Configuration Change Prediction

We used Machine Learning (ML) algorithms on data provided by a client’s real network to identify the role of optical device metrics in detecting flapping links (links that go down multiple times a day) and Interior Gateway Protocol (IGP) configuration changes in the next five days based on data collected for the previous five days. Our prototype shows that we could predict upcoming IGP changes five days ahead with 95% precision and 75% recall. Adding optical data to the ML model increased the performance by about 9%. The importance of optical features like “OCH-OPTMAX”, “OCH-OPTMIN”, “OCH-DGDMAX”, “E-SES”, “E-UAS” and “E-ES” that account for optical minimum power, unavailable seconds, and errored seconds is confirmed using common feature importance wrapper methods. This ML approach is significantly better than existing manual methods followed by the operator.