Table 5 Comparison of performance of our multi-layer approach vs the baseline (single-layer). We show AUROC values for the LR classifier evaluated on different size classes of both US and IT datasets
From: A multi-layer approach to disinformation detection in US and Italian news spreading on Twitter
Size class | Single-layer | Multi-layer |
|---|---|---|
(US) [0,100) | 0.74 ± 0.02 | 0.87 ± 0.01 |
(US) [100,1000) | 0.85 ± 0.02 | 0.93 ± 0.01 |
(US) [1000,+∞) | 0.93 ± 0.03 | 0.94 ± 0.02 |
(US) [0,+∞) | 0.78 ± 0.02 | 0.88 ± 0.01 |
(IT) [0,100) | 0.77 ± 0.08 | 0.89 ± 0.06 |
(IT) [100,1000) | 0.66 ± 0.14 | 0.86 ± 0.07 |
(IT) [0,+∞) | 0.74 ± 0.12 | 0.90 ± 0.02 |