Spatio-Temporal Graph Attention Networks for Alarm Root Cause Recognition in Optical Transport Network

Conference: ECOC 2024 - 50th European Conference on Optical Communication
09/22/2024 - 09/26/2024 at Frankfurt, Germany

Proceedings: ITG-Fb. 317: ECOC 2024

Pages: 4Language: englishTyp: PDF

Authors:
Yang, Weijie; Zhang, Chunyu; Jiang, Xunjie; Fan, Yanlin; Bi, Zhongbo; Xiong, Jiansheng; Xiao, Xue; Zhang, Min; Wang, Danshi

Abstract:
We propose a root alarm recognition scheme based on graph attention network, which can capture the spatio-temporal information of alarms. The F1-score for root alarm recognition in real backbone network is 0.9567, and the output spatio-temporal subgraph can provide interpretable support for recognition results.