EUVIMEDEUVIMED entdecken
Crossref

FedSTGAT: Federated Spatio-Temporal Graph Attention Network for Root Alarm Identification in Optical Transport Network

Weijie Yang, Chunyu Zhang, Cheng Xing, Min Zhang, Xunjie Jiang, Yanlin Fan, Zhongbo Bi, Jiansheng Xiong, Danshi Wang

IEEE Transactions on Cognitive Communications and Networking · 2026 · 12 · 8679-8690

DOI: 10.1109/tccn.2026.3694892 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Cognitive Communications and Networking, 2026. Verfasst von Weijie Yang und 8 weiteren Autor:innen. Die Arbeit nennt 36 Literaturangaben. Gefördert von: National Natural Science Foundation of China, National Key Research and Development Program of China, BUPT Excellent Ph.D. Students Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In Optical Transport Networks (OTNs), faults often generate numerous derived alarms that may obscure the true root alarm. When multiple faults occur within the same time window, their alarm propagation may overlap, so the earliest observed alarm is not necessarily the root alarm. Accurate Root Alarm Identification (RAI) is therefore essential for timely fault management and reliable network operation. Traditional RAI methods rely on handcrafted alarm rules and struggle to capture the spatio-temporal nature of alarm propagation. Existing spatio-temporal methods, however, often focus on a single modality or localized dependencies and therefore overlook the coupled interactions among topology, time, and propagation behavior, which limits their accuracy and scalability in large OTNs. To address these challenges, we propose a Federated Spatio-Temporal Graph Attention Network (FedSTGAT). FedSTGAT transforms raw alarm records into spatio-temporal subgraphs and applies STGAT for RAI. We further incorporate a meta-path-consistent learning objective to distill higher-order propagation patterns, improving both accuracy and interpretability. To address data silos and privacy constraints in multi-domain OTNs, we integrate FedSTGAT into a federated learning framework that enables collaborative training without sharing raw alarms. Experimental results show that FedSTGAT achieves an RAI accuracy of 0.9653, improving accuracy by 0.2473 on average over rule-based methods, and remains above 0.91 in the federated setting. Overall, FedSTGAT provides an effective and interpretable framework for intelligent RAI to support fault diagnosis and operational decision-making.

Förderung

  • National Natural Science Foundation of China
  • National Key Research and Development Program of China
  • BUPT Excellent Ph.D. Students Foundation

Zitieren

APA

Yang, W., Zhang, C., Xing, C., Zhang, M., Jiang, X., Fan, Y., Bi, Z., Xiong, J., & Wang, D. (2026). FedSTGAT: Federated Spatio-Temporal Graph Attention Network for Root Alarm Identification in Optical Transport Network. IEEE Transactions on Cognitive Communications and Networking, 12, 8679-8690. https://doi.org/10.1109/tccn.2026.3694892

Vancouver

Yang W, Zhang C, Xing C, Zhang M, Jiang X, Fan Y, et al. FedSTGAT: Federated Spatio-Temporal Graph Attention Network for Root Alarm Identification in Optical Transport Network. IEEE Transactions on Cognitive Communications and Networking. 2026;12:8679-8690. doi:10.1109/tccn.2026.3694892

BibTeX
@article{Yang2026,
  author = {Yang, Weijie and Zhang, Chunyu and Xing, Cheng and Zhang, Min and Jiang, Xunjie and Fan, Yanlin and Bi, Zhongbo and Xiong, Jiansheng and Wang, Danshi},
  title = {FedSTGAT: Federated Spatio-Temporal Graph Attention Network for Root Alarm Identification in Optical Transport Network},
  journal = {IEEE Transactions on Cognitive Communications and Networking},
  year = {2026},
  volume = {12},
  pages = {8679--8690},
  doi = {10.1109/tccn.2026.3694892}
}
Kostenlos starten →