Skip to main content

Table 5 Cross-region model evaluation for vax-skeptic content prediction with respect to different modalities

From: Network embedding aided vaccine skepticism detection

 

EU

US

Diff (%)

(a) Only text

 EU

0.775

0.771

0.5

 US

0.711

0.745

4.7

(b) Text + user history

 EU

0.836

0.826

1.2

 US

0.783

0.811

3.6

(c) Text + network stats

 EU

0.842

0.780

7.9

 US

0.717

0.769

7.2

(d) Text + raw network

 EU

0.885

0.874

1.2

 US

0.803

0.846

5.3

(e) All

 EU

0.888

0.876

1.4

 US

0.832

0.847

1.8

  1. In each table marked by letters (a–e), columns and rows represent the training and test regions, respectively. Here, raw network represents Walklets (Perozzi et al. 2016), the best-performing node embedding model (see Fig. 11). The diff column shows the AUC difference between training and testing on the same domain versus out-of-domain