Skip to main content

Table 11 PUMD and multiple comparison algorithms performance evaluation

From: PUMD: a PU learning-based malicious domain detection framework

Label frequency

PUMD

PU_biased

PU_empirical

mix_supervised

Pure_supervised

Unsupervised

Improve*

F1

MCC

F1

MCC

F1

MCC

F1

MCC

F1

MCC

F1

MCC

F1

MCC

0.1

0.9326

0.9297

0.8932

0.8895

0.9105

0.9066

0.9163

0.9137

0.9115

0.9091

0.5330

0.5131

0.0163

0.0160

0.2

0.9379

0.9357

0.9092

0.9058

0.9036

0.9001

0.9001

0.8984

0.8841

0.8831

0.5578

0.5479

0.0378

0.0373

0.3

0.9284

0.9262

0.9143

0.9114

0.9101

0.9078

0.8974

0.8965

0.8952

0.8945

0.4695

0.4566

0.0310

0.0297

0.4

0.9216

0.9198

0.9141

0.9118

0.8998

0.8980

0.8951

0.8941

0.8989

0.8983

0.4959

0.4816

0.0266

0.0257

0.5

0.9208

0.9199

0.9028

0.9012

0.8906

0.8899

0.8555

0.8578

0.8603

0.8628

0.4204

0.4223

0.0652

0.0621

0.6

0.8975

0.8971

0.8817

0.8811

0.8704

0.8707

0.8377

0.8422

0.8320

0.8370

0.3953

0.4009

0.0598

0.0549

0.7

0.8805

0.8816

0.8608

0.8631

0.8491

0.8526

0.7675

0.7820

0.7786

0.7925

0.4122

0.4096

0.1130

0.0996

0.8

0.8485

0.8541

0.8108

0.8192

0.7965

0.8066

0.7592

0.7790

0.7724

0.7894

0.3882

0.3849

0.0893

0.0750

0.9

0.7542

0.7736

0.6926

0.7226

0.6820

0.7140

0.5863

0.6398

0.5928

0.6470

0.3390

0.3444

0.1679

0.1338

  1. Underlined value: overall best method, *count as Score (PUMD) \(-\) Score (mix_supervised(baseline))