A graph neural architecture search approach for identifying bots in social media
2024

Detecting Bots on Social Media Using Graph Neural Networks

Sample size: 229580 publication 10 minutes Evidence: high

Author Information

Author(s): Tzoumanekas Georgios, Chatzianastasis Michail, Ilias Loukas, Kiokes George, Psarras John, Askounis Dimitris

Primary Institution: National Technical University of Athens

Hypothesis

Can a Neural Architecture Search approach improve bot detection in social media by leveraging graph structures?

Conclusion

The proposed model achieved an accuracy of 85.68%, surpassing state-of-the-art models in bot detection.

Supporting Evidence

  • The model constructed a graph with 229,580 nodes and 227,979 edges.
  • DFG-NAS was adapted to automatically search for optimal configurations in the bot detection task.
  • The model achieved an accuracy of 85.7%, surpassing existing models.
  • User metadata was integrated into the graph to enhance detection capabilities.
  • An ablation study confirmed the importance of using all features for optimal performance.

Takeaway

This study created a smart computer program that can find fake accounts on social media by looking at how users are connected to each other.

Methodology

The study used a Deep and Flexible Graph Neural Architecture Search (DFG-NAS) to optimize bot detection through Relational Graph Convolutional Networks (RGCNs) on the TwiBot-20 dataset.

Potential Biases

The model's performance may be affected by the quality and representativeness of the labeled data used for training.

Limitations

The study was conducted on a single dataset, which may limit the generalizability of the findings.

Participant Demographics

The dataset includes user profiles from the TwiBot-20 dataset, which consists of both bot and human accounts.

Statistical Information

P-Value

0.004

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.3389/frai.2024.1509179

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication