Skip to main content

Table 2 Classification performance comparisons on the UCD dataset in the frequency domain

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

 

UCD Dataset (frequency Domain)

Method

AUC-ROC

AUC-PR

Macro-F1

Micro-F1

Accuracy

PSVM

0.8177 ±0.0142

0.5767 ±0.0172

0.5204 ±0.0275

0.5854 ±0.0733

0.6193 ±0.1053

DNN

0.7213 ±0.1435

0.5224 ±0.1048

0.3542 ±0.2171

0.4331 ±0.2269

0.5262 ±0.1613

RNN

0.6228 ±0.0465

0.3350 ±0.0394

0.2663 ±0.0241

0.3970 ±0.0428

0.5091 ±0.0391

RNNAtt l

0.6172 ±0.0386

0.3305 ±0.0386

0.2457 ±0.0307

0.3734 ±0.0566

0.5002 ±0.0476

RNNAtt c

0.6234 ±0.0451

0.3335 ±0.0345

0.2554 ±0.0258

0.3712 ±0.0325

0.5010 ±0.0367

CNN

0.8732 ±0.0129

0.6725 ±0.0120

0.5925 ±0.0604

0.6492 ±0.0841

0.6590 ±0.0979

CRNN

0.8660 ±0.0074

0.6454 ±0.0135

0.5693 ±0.0060

0.6395 ±0.0370

0.6634 ±0.0412

CRNNAtt l

0.8570 ±0.0183

0.6281 ±0.0359

0.5810 ±0.0371

0.6486 ±0.0641

0.6683 ±0.0657

CRNNAtt c

0.8671 ±0.0274

0.6418 ±0.0401

0.5849 ±0.0577

0.6528 ±0.0547

0.6791 ±0.0546

ChannelAtt

0.8705 ±0.0483

0.6818 ±0.0580

0.6517 ±0.0334

0.7070 ±0.0605

0.7152 ±0.0574

HybridAtt l

0.8719 ±0.0214

0.6669 ±0.0297

0.6342 ±0.0316

0.6962 ±0.0645

0.7070 ±0.0707

HybridAtt f

0.8854 ±0.0137

0.6886 ±0.0256

0.6639 ±0.0301

0.7231 ±0.0489

0.7328 ±0.0546