An Enhanced Misinformation Detection Model Based on an Improved Beluga Whale Optimization Algorithm and Cross-Modal Feature Fusion DOI Creative Commons
Guangyu Mu,

Xingwang Ju,

Huibin Yan

et al.

Biomimetics, Journal Year: 2025, Volume and Issue: 10(3), P. 128 - 128

Published: Feb. 20, 2025

The proliferation of multimodal misinformation on social media has become a critical concern. Although detection methods have advanced, feature representation and cross-modal semantic alignment challenges continue to hinder the effective use data. Therefore, this paper proposes an IBWO-CASC model that integrates improved Beluga Whale Optimization algorithm with attention fusion. Firstly, is enhanced by combining adaptive search mechanisms batch parallel strategies in space. Secondly, method designed based supervised contrastive learning establish consistency. Then, incorporates Cross-modal Attention Promotion mechanism global–local interaction pattern. Finally, multi-task framework built classification objectives. empirical analysis shows proposed achieves accuracy 97.41% our self-constructed dataset. Compared average existing six baseline models, 4.09%. Additionally, it demonstrates robustness handling complex scenarios.

Language: Английский

An Enhanced Misinformation Detection Model Based on an Improved Beluga Whale Optimization Algorithm and Cross-Modal Feature Fusion DOI Creative Commons
Guangyu Mu,

Xingwang Ju,

Huibin Yan

et al.

Biomimetics, Journal Year: 2025, Volume and Issue: 10(3), P. 128 - 128

Published: Feb. 20, 2025

The proliferation of multimodal misinformation on social media has become a critical concern. Although detection methods have advanced, feature representation and cross-modal semantic alignment challenges continue to hinder the effective use data. Therefore, this paper proposes an IBWO-CASC model that integrates improved Beluga Whale Optimization algorithm with attention fusion. Firstly, is enhanced by combining adaptive search mechanisms batch parallel strategies in space. Secondly, method designed based supervised contrastive learning establish consistency. Then, incorporates Cross-modal Attention Promotion mechanism global–local interaction pattern. Finally, multi-task framework built classification objectives. empirical analysis shows proposed achieves accuracy 97.41% our self-constructed dataset. Compared average existing six baseline models, 4.09%. Additionally, it demonstrates robustness handling complex scenarios.

Language: Английский

Citations

0