Table 7 Results of comparison among different variants of KAGN on twitter datasets
From: KAGN:knowledge-powered attention and graph convolutional networks for social media rumor detection
Dataset | Variant | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|
Twitter15 | PTE | 0.8036 | 0.8053 | 0.8036 | 0.8037 |
PTE + KGE | 0.8406 | 0.8486 | 0.8408 | 0.8417 | |
PTE + KAE | 0.8636 | 0.8690 | 0.8628 | 0.8642 | |
KAGN | 0.8923 | 0.8947 | 0.8905 | 0.8956 | |
Twitter16 | PTE | 0.8295 | 0.8364 | 0.8296 | 0.8315 |
PTE + KGE | 0.8672 | 0.8734 | 0.8676 | 0.8650 | |
PTE + KAE | 0.8750 | 0.8749 | 0.8742 | 0.8744 | |
KAGN | 0.9013 | 0.9034 | 0.9062 | 0.8976 | |
PHEME | PTE | 0.8125 | 0.8056 | 0.8125 | 0.8052 |
PTE + KGE | 0.8281 | 0.8228 | 0.8281 | 0.8208 | |
PTE + KAE | 0.8594 | 0.8594 | 0.8594 | 0.8594 | |
KAGN | 0.8646 | 0.8402 | 0.8293 | 0.8344 | |
Politifact | PTE | 0.8203 | 0.8203 | 0.8210 | 0.8202 |
PTE + KGE | 0.8438 | 0.8434 | 0.8434 | 0.8438 | |
PTE + KAE | 0.8672 | 0.8683 | 0.8672 | 0.8673 | |
KAGN | 0.8790 | 0.8768 | 0.8780 | 0.8751 |