Table 1 From a purely empirical perspective, the datasets considered in the present contribution are characterised by a small link density and different percentages of positive and negative links: the connectance of U.S. Senate and U.S. House of Representatives amounts to 0.1 while 55% of links is positive; the connectance of FilmTrust amounts to 0.01 while 80% of links is positive

From: Statistically validated projection of bipartite signed networks

 

FilmTrust

U.S. Senate

U.S. House of representatives

 

L

L+

c

L

L+

c

L

L+

c

Zero-deflated projection—naive

137.601

725.983

0.76

1.379

2.329

0.36

18.512

27.922

0.35

Zero-deflated projection—global filter

27.940

343.181

0.33

689

1.598

0.22

11.681

21.419

0.25

Zero-deflated projection—local filter

116.583

353.224

0.41

1.220

1.312

0.24

16.323

16.897

0.25

Zero-deflated projection—naive

0

1.135.278

1

34

10.406

1

14.379

117.976

1

Zero-inflated projection—global filter

77.024

667.448

0.65

3.942

5.041

0.86

47.839

67.631

0.87

Zero-inflated projection—local filter

56.803

62.057

0.10

2.897

2.337

0.50

40.982

32.693

0.56

  1. Projections, instead, are characterised by a link density that depends on the scheme employed for filtering: overall, however, it is smaller than that of the naive projections, the local filter cutting more edges than the global one.