Transformer-based Alarm Context-Vectorization Representation for Reliable Alarm Root Cause Identification in Optical Networks

Abstract

A Transformer-based alarm context-vectorization representation technique is proposed for alarm root cause identification and correlation analysis. Three common root alarms are identified with an accuracy of 94.47%, and other correlated alarms are obtained with quantified correlation degrees.