EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Explainable Root Cause Analysis for Unsupervised Soft Failure Detection in Optical Networks

Yuxiao Wang, Emilio Paolini, Azarm Yeganehfallah, Andrea Sgambelluri, Sourav Mondal, Ye Pu, Luca Valcarenghi, Elaine Wong

IEEE Open Journal of the Communications Society · 2026 · 7 · 6952-6963

DOI: 10.1109/ojcoms.2026.3704961 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Open Journal of the Communications Society, 2026. Verfasst von Yuxiao Wang und 7 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Faculty of Engineering and Information Technology Visiting Fellows Scheme, The University of Melbourne.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The increasing scale of optical transport networks and the widespread deployment of optical performance monitoring have enabled the collection of high-dimensional telemetry data, driving extensive research on machine-learning-based failure detection and localization. However, labeled failure data are usually scarce in practical operation, and failure patterns drift over time, causing supervised learning methods to struggle with adapting to new failure types. Although unsupervised learning can detect anomalous samples in the absence of labels, the model outputs often lack semantic meaning, thereby preventing the outputs from mapping to specific failure types. To address this problem, we propose a framework for soft failure detection and root cause analysis guided by explainable artificial intelligence. Firstly, an online anomaly detector and an unsupervised K-means model group unlabeled telemetry samples into normal or failure clusters. Secondly, the explainable root cause analysis module maps these non-semantic clusters to interpretable failure labels. Experimental results show that the framework achieves up to 96.6% cluster-to-label mapping accuracy in providing failure labels, without requiring labeled failure data for training.

Förderung

  • Faculty of Engineering and Information Technology Visiting Fellows Scheme, The University of Melbourne

Zitieren

APA

Wang, Y., Paolini, E., Yeganehfallah, A., Sgambelluri, A., Mondal, S., Pu, Y., Valcarenghi, L., & Wong, E. (2026). Explainable Root Cause Analysis for Unsupervised Soft Failure Detection in Optical Networks. IEEE Open Journal of the Communications Society, 7, 6952-6963. https://doi.org/10.1109/ojcoms.2026.3704961

Vancouver

Wang Y, Paolini E, Yeganehfallah A, Sgambelluri A, Mondal S, Pu Y, et al. Explainable Root Cause Analysis for Unsupervised Soft Failure Detection in Optical Networks. IEEE Open Journal of the Communications Society. 2026;7:6952-6963. doi:10.1109/ojcoms.2026.3704961

BibTeX
@article{Wang2026,
  author = {Wang, Yuxiao and Paolini, Emilio and Yeganehfallah, Azarm and Sgambelluri, Andrea and Mondal, Sourav and Pu, Ye and Valcarenghi, Luca and Wong, Elaine},
  title = {Explainable Root Cause Analysis for Unsupervised Soft Failure Detection in Optical Networks},
  journal = {IEEE Open Journal of the Communications Society},
  year = {2026},
  volume = {7},
  pages = {6952--6963},
  doi = {10.1109/ojcoms.2026.3704961}
}
Kostenlos starten →