Table 4 Comparison of FSCL with Different Schemes.

From: Intelligent diagnosis of gearbox in data heterogeneous environments based on federated supervised contrastive learning framework

Datasets

DDS

WTDS

Schemes

Metrics

Methods

Central

FedAvg

FedProx

FSCL

Central

FedAvg

FedProx

FSCL

IID

Accuracy

0.9585

0.9699

0.9738

0.9936

0.9067

0.9195

0.9451

0.9507

Specificity

0.9594

0.9802

0.9840

0.9958

0.9039

0.9229

0.9425

0.9574

Sensitivity

0.7259

0.8878

0.8923

0.9764

0.5489

0.8922

0.9656

0.8978

AUC

0.9664

0.9907

0.9937

0.9995

0.9108

0.9567

0.9734

0.9741

F1

0.6897

0.8711

0.8862

0.9718

0.4959

0.7308

0.8231

0.8472

Non-

IID- Class

Accuracy

0.9140

0.9825

0.9842

0.9906

0.8722

0.8936

0.9205

0.9322

Specificity

0.9101

0.9888

0.9899

0.9937

0.8401

0.9384

0.9534

0.9353

Sensitivity

0.6733

0.9327

0.9383

0.9501

0.5522

0.5031

0.5914

0.8989

AUC

0.9315

0.9970

0.9990

0.9993

0.8699

0.9093

0.9549

0.9631

F1

0.5840

0.9229

0.9313

0.9516

0.4428

0.4358

0.5853

0.7608

Non-

IID- Client

Accuracy

-

0.9277

0.9323

0.9593

-

0.9089

0.9184

0.9207

Specificity

0.9731

0.9762

0.9738

0.9164

0.9288

0.9493

Sensitivity

0.7396

0.7419

0.8205

0.8489

0.8567

0.8980

AUC

0.9553

0.9532

0.9860

0.9604

0.9633

0.9631

F1

0.8002

0.8240

0.8206

0.6833

0.7161

0.8470

Non-

IID- Domain

Accuracy

0.8328

0.9672

0.9779

0.9892

0.7589

0.9064

0.9156

0.9383

Specificity

0.9066

0.9753

0.9823

0.9924

0.8591

0.9357

0.9417

0.9356

Sensitivity

0.6144

0.9024

0.9428

0.9630

0.4711

0.6722

0.7067

0.9011

AUC

0.9258

0.9889

0.9937

0.9991

0.6249

0.9525

0.9594

0.9679

F1

0.5111

0.8676

0.9103

0.9534

0.4085

0.5898

0.6394

0.7845

Back to article page