Table 6 Performance of our implemented models under a unified experimental setup, including BERT and RoBERTa baselines.

From: Graph-augmented transformer ensemble framework for robust and scalable fake news detection in social media ecosystems

Model

Accuracy (%)

Precision (%)

Recall (%)

F1-Score (%)

ROC-AUC(%)

SVM(TF-IDF)

85.2

84.7

85.5

85.1

87.2

BERT

90.2

89.8

90.4

90.1

91.6

RoBERTa

91.3

90.7

91.8

91.2

92.4

GCN(User-Post)

89.5

88.3

89.8

89.0

91.0

GAT

90.0

89.4

90.1

89.7

91.5

CSI

88.4

87.2

89.1

88.1

90.1

Proposed Model

96.3

95.9

96.6

96.2

97.1

Back to article page