Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Graph neural networks based log anomaly detection and explanation.

Zhong Li1,2, Jiayang Shi1, Matthijs van Leeuwen1

  • 1Leiden Institute of Advanced Computer Science (LIACS), Leiden University, Leiden, The Netherlands.

Data Mining and Knowledge Discovery
|July 6, 2026
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

The added value of ferritin levels and genetic markers for the prediction of haemoglobin deferral.

Vox sanguinis·2023
Same author

Generating synthetic mixed discrete-continuous health records with mixed sum-product networks.

Journal of the American Medical Informatics Association : JAMIA·2022
Same author

Explainable haemoglobin deferral predictions using machine learning models: Interpretation and consequences for the blood supply.

Vox sanguinis·2022
Same author

Choosing and changing the analysis scale in non-inferiority trials with a binary outcome.

Clinical trials (London, England)·2021
Same journal

Topology only pre-training: towards generalised multi-domain graph models.

Data mining and knowledge discovery·2026
Same journal

Detection and evaluation of clusters within sequential data.

Data mining and knowledge discovery·2025
Same journal

Universal representation learning for multivariate time series using the instance-level and cluster-level supervised contrastive learning.

Data mining and knowledge discovery·2025
Same journal

Missing value replacement in strings and applications.

Data mining and knowledge discovery·2025
Same journal

Robust explainer recommendation for time series classification.

Data mining and knowledge discovery·2024
Same journal

Somtimes: self organizing maps for time series clustering and its application to serious illness conversations.

Data mining and knowledge discovery·2024
See all related articles

This study introduces Logs2Graphs, a novel graph-based method for unsupervised log anomaly detection. It achieves high accuracy by converting logs into graphs and using graph neural networks, offering explainable insights for system monitoring.

Area of Science:

  • Computer Science
  • Artificial Intelligence
  • Machine Learning

Background:

  • Event logs are crucial for monitoring high-tech systems.
  • Existing log anomaly detection methods often lack accuracy due to limited input analysis.
  • Quantitative or sequential analysis alone is insufficient for robust anomaly detection.

Purpose of the Study:

  • To propose a novel graph-based method for unsupervised log anomaly detection.
  • To improve the accuracy and explainability of log anomaly detection systems.
  • To introduce a new graph neural network model for detecting anomalies in attributed, directed, and weighted graphs.

Main Methods:

  • Developed Logs2Graphs: a method converting event logs into attributed, directed, and weighted graphs.
  • Introduced One-Class Digraph Inception Convolutional Networks (OCDiGCN), a graph neural network for graph-level anomaly detection.
Keywords:
Graph neural networksKnowledge discovery from graphsLog analysisLog anomaly detection

Related Experiment Videos

  • Integrated graph representation learning with anomaly detection for specialized representations.

    • Main Results:

    • Logs2Graphs demonstrated comparable or superior performance against state-of-the-art methods on five benchmark datasets.
    • The OCDiGCN model achieved high detection accuracy by learning specialized graph representations.
    • The method provides explanations for detected anomalies by identifying pivotal nodes.

    Conclusions:

    • Graph-based approaches, like Logs2Graphs, offer a powerful alternative for unsupervised log anomaly detection.
    • OCDiGCN effectively detects graph-level anomalies and enhances system monitoring.
    • The explainability feature aids in root cause analysis for system failures.