Microblog-HAN: A micro-blog rumor detection model based on heterogeneous graph attention network

Area of Science:

• Computer Science
• Social Media Analysis
• Artificial Intelligence

Background:

• Social media facilitates information spread but also amplifies Internet rumors.
• Early detection of rumors is crucial for mitigating their impact.
• Existing methods often overlook the semantic richness of microblog propagation graphs.

Purpose of the Study:

• To develop an advanced rumor detection model that leverages microblog information propagation graph semantics.
• To propose a novel graph-based approach for early-stage rumor identification.

Main Methods:

• Modeling microblog information transmission as a heterogeneous graph.
• Constructing Microblog-HAN, a graph neural network incorporating attention mechanisms.
• Extracting textual and visual features, employing node-level and semantic-level attention for embedding generation.

Main Results:

• The Microblog-HAN model achieved over 92% accuracy in detecting microblog rumors.
• Demonstrated superior performance compared to existing methods on real-world datasets (Weibo2016, Weibo2021).
• Effectively captured and aggregated semantic information from the propagation graph.

Conclusions:

• Microblog-HAN offers a robust solution for early rumor detection in social media.
• The model's ability to utilize graph semantics significantly enhances detection accuracy.
• This approach provides a promising direction for combating online misinformation.