Table 4 Evaluation indicator scores for different sampling methods and sampling rates of test data.

From: A modified generative adversarial networks method for assisting the diagnosis of deep venous thrombosis complications in stroke patients

Performance

SR = 0.5

SR = 0.75

SR = 1

AUC

G

F1

AUC

G

F1

AUC

G

F1

RUS

0.6570

0.5823

0.4750

0.6119

0.5323

0.4837

0.6377

0.5835

0.5945

ENN

0.6739

0.0931

0.0487

0.6739

0.0931

0.0487

0.6739

0.0931

0.0487

Tomek Links

0.6490

0.0763

0.0515

0.6490

0.0763

0.0515

0.6490

0.0763

0.0515

ROS

0.8318

0.7010

0.6178

0.8389

0.7441

0.7069

0.8397

0.7786

0.7888

ADASYN

0.9113

0.8194

0.7700

0.9296

0.8671

0.8497

0.9357

0.8801

0.8837

SMOTE

0.9091

0.8298

0.7828

0.9277

0.8645

0.8457

0.9353

0.8787

0.8810

BS-1

0.9188

0.8451

0.7998

0.9322

0.8715

0.8536

0.9405

0.8788

0.8808

BS-2

0.8994

0.8287

0.7783

0.9165

0.8569

0.8367

0.9226

0.8622

0.8658

SMOTE+ENN

0.9233

0.8642

0.8518

0.9342

0.8720

0.8968

0.9342

0.8739

0.8746

SMOTE+Tomek

0.9086

0.8305

0.7820

0.9269

0.8688

0.8508

0.9351

0.8788

0.8808

BS-1+Tomek

0.9195

0.8436

0.7984

0.9324

0.8680

0.8496

0.9416

0.8820

0.8834

BS-2+Tomek

0.8999

0.8263

0.7754

0.9178

0.8554

0.8357

0.9235

0.8664

0.8693

WGAN

0.9470

0.9221

0.9269

0.9643

0.9492

0.9511

0.9706

0.9619

0.9626

ACWGAN

0.9494

0.9222

0.9269

0.9664

0.9497

0.9515

0.9757

0.9627

0.9635