Optical Network Diagnostics Using Graph Neural Networks and Natural Language Processing

We propose an AI-powered network diagnostic strategy including alarm clustering and fault localization with >98% accuracy for up-1o 16-degree ROADMs and demonstrate the advantage of using NLP in encoding.