Skip to main content

Table 3 Classification performance comparisons on the UCD dataset in the time domain

From: A hybrid self-attention deep learning framework for multivariate sleep stage classification

 

UCD Dataset (Time Domain)

Method

AUC-ROC

AUC-PR

Macro-F1

Micro-F1

Accuracy

PSVM

0.4945 ±0.0068

0.2249 ±0.0042

0.1352 ±0.0388

0.2584 ±0.0464

0.3877 ±0.1107

DNN

0.5024 ±0.0025

0.5128 ±0.0144

0.1129 ±0.0282

0.2372 ±0.1023

0.3962 ±0.1193

RNN

0.6236 ±0.0424

0.3352 ±0.0358

0.2464 ±0.0360

0.3417 ±0.0908

0.4487 ±0.1174

RNNAtt l

0.6256 ±0.0297

0.3254 ±0.0279

0.2557 ±0.0328

0.3483 ±0.0938

0.4521 ±0.1100

RNNAtt c

0.6279 ±0.0549

0.3434 ±0.0376

0.2465 ±0.0314

0.3328 ±0.0870

0.4501 ±0.1087

CNN

0.8421 ±0.0186

0.5844 ±0.0300

0.5775 ±0.0336

0.6493 ±0.0326

0.6595 ±0.0347

CRNN

0.8453 ±0.0229

0.5945 ±0.0297

0.5761 ±0.0345

0.6483 ±0.0383

0.6592 ±0.0456

CRNNAtt l

0.8461 ±0.0115

0.6097 ±0.0206

0.5954 ±0.0362

0.6585 ±0.0437

0.6659 ±0.0473

CRNNAtt c

0.8505 ±0.0140

0.6120 ±0.0273

0.6004 ±0.0319

0.6622 ±0.0509

0.6632 ±0.0541

ChannelAtt

0.8720 ±0.0203

0.6834 ±0.0278

0.6107 ±0.0324

0.6907 ±0.0545

0.7169 ±0.0624

HybridAtt l

0.8885 ±0.0142

0.7009 ±0.0223

0.6689 ±0.0314

0.7264 ±0.0491

0.7317 ±0.0512

HybridAtt f

0.8966 ±0.0214

0.7082 ±0.0283

0.6818 ±0.0304

0.7368 ±0.0591

0.7424 ±0.0594