Skip to main content

Table 2 Edge- and User-level Performances in Exit_DS

From: Unveiling the silent majority: stance detection and characterization of passive users on social media using collaborative filtering and graph convolutional networks

Models

Edge

User

nDCG

MAP

Acc.

RMSE

Least

Middle

Most

Null Model

0.0358

0.0094

0.5596

0.6655

0.4583

0.5692

0.6514

CSMF

0.0557

0.0404

0.3955

0.7775

0.4211

0.3857

0.3806

NGCF

0.0728

0.0335

0.6312

0.6132

0.4833

0.5846

0.8257

LightGCN

0.0773

0.0388

0.6599

0.5766

0.5462

0.5956

0.8380

WLGCN (norm by user)

0.0754

0.0385

0.6543

0.5948

0.5083

0.5923

0.8624

WLGCN (norm by TF-IDF)

0.0776

0.0397

0.6494

0.5994

0.5333

0.5615

0.8532

WLGCN

0.0783

0.0395+

0.6642+

0.5860

0.5646+

0.6015

0.8267

WLGCN + (\(E_{\mathrm{HT}}\))

0.0885+

0.0449

0.6702+

0.5743+

0.5622+

0.6150

0.8332

WLGCN + (\(\mathcal{G}_{\mathrm{PathSim}}\))

0.0769+

0.0392

0.6795

0.5658

0.5757+

0.6276

0.8352

WLGCN + (\(E_{\mathrm{HT}}\), \(\mathcal{G}_{\mathrm{PathSim}}\))

0.0902

0.0458

0.7065

0.5422

0.5733+

0.6654

0.8807

WLGCN + (\(\mathcal{G}_{\mathrm{Soc}}\))

0.0770+

0.0390

0.6896

0.5578

\(\textbf{0.5900}^{+}\)

0.6346

0.8440

WLGCN + (\(E_{\mathrm{HT}}\), \(\mathcal{G}_{\mathrm{Soc}}\))

\(\textbf{0.0908}^{*}\)

\(\textbf{0.0459}^{*}\)

\(\textbf{0.7118}^{*}\)

\(\textbf{0.5367}^{*}\)

0.5817

0.6731

0.8807

WLGCN + (\(\mathcal{G}_{\mathrm{PathSim}}\), \(\mathcal{G}_{\mathrm{Soc}}\))

0.0749+

0.0384

0.6922

0.5496

0.5567

0.6577

0.8624

WLGCN + (\(E_{\mathrm{HT}}\), \(\mathcal{G}_{\mathrm{PathSim}}\), \(\mathcal{G}_{\mathrm{Soc}}\))

0.0881

0.0444

0.7091

0.5402+

0.5567

\(\textbf{0.6808}^{*}\)

\(\textbf{0.8899}^{*}\)

  1. We marked WLGCN variants’ results that are statistically significant (t-test) compared to the baseline LightGCN: (p<0.05), +(p<0.1).