Table 4 The average predicting AUC obtained by 100 independent runs on 17 real directed networks.

From: Link prediction in complex networks via matrix perturbation and decomposition

Network

CN

AA

RA

SPM

LR

MPD

Bison

0.798

0.801

0.794

0.714

0.770(0.16)

0.779(0.09)

Cattle

0.809

0.798

0.802

0.517

0.676(0.16)

0.800(0.14)

Football

0.966

0.964

0.966

0.791

0.725(0.22)

0.948(0.11)

Gramdry

0.750

0.761

0.757

0.757

0.924(0.14)

0.956(0.13)

Gramwet

0.752

0.763

0.760

0.748

0.935(0.14)

0.963(0.15)

Cypdry

0.783

0.781

0.780

0.822

0.868(0.14)

0.951(0.13)

Cypwet

0.783

0.783

0.779

0.832

0.867(0.14)

0.952(0.11)

Mangdry

0.761

0.775

0.783

0.771

0.887(0.12)

0.945(0.12)

Mangwet

0.769

0.783

0.787

0.785

0.893(0.12)

0.951(0.09)

Polbooks

0.903

0.891

0.903

0.876

0.553(0.22)

0.918(0.38)

Baydry

0.739

0.740

0.742

0.821

0.898(0.11)

0.965(0.09)

Baywet

0.742

0.746

0.751

0.822

0.902(0.12)

0.956(0.10)

C. elegans

0.805

0.809

0.812

0.902

0.593(0.16)

0.903(0.07)

USAir

0.966

0.964

0.973

0.955

0.833(0.10)

0.973(0.22)

Email-Eu

0.946

0.948

0.952

0.966

0.749(0.06)

0.962(0.02)

PB

0.928

0.916

0.927

0.943

0.669(0.04)

0.961(0.02)

CollegeMsg

0.735

0.747

0.755

0.960

0.546(0.24)

0.921(0.01)

  1. The training set contains 90% of total connections. The values in the brackets are the values of optimal parameters of the methods. The highest precisions are emphasized by boldface.