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)
MethodAUC-ROCAUC-PRMacro-F1Micro-F1Accuracy
PSVM0.4945 ±0.00680.2249 ±0.00420.1352 ±0.03880.2584 ±0.04640.3877 ±0.1107
DNN0.5024 ±0.00250.5128 ±0.01440.1129 ±0.02820.2372 ±0.10230.3962 ±0.1193
RNN0.6236 ±0.04240.3352 ±0.03580.2464 ±0.03600.3417 ±0.09080.4487 ±0.1174
RNNAtt l0.6256 ±0.02970.3254 ±0.02790.2557 ±0.03280.3483 ±0.09380.4521 ±0.1100
RNNAtt c0.6279 ±0.05490.3434 ±0.03760.2465 ±0.03140.3328 ±0.08700.4501 ±0.1087
CNN0.8421 ±0.01860.5844 ±0.03000.5775 ±0.03360.6493 ±0.03260.6595 ±0.0347
CRNN0.8453 ±0.02290.5945 ±0.02970.5761 ±0.03450.6483 ±0.03830.6592 ±0.0456
CRNNAtt l0.8461 ±0.01150.6097 ±0.02060.5954 ±0.03620.6585 ±0.04370.6659 ±0.0473
CRNNAtt c0.8505 ±0.01400.6120 ±0.02730.6004 ±0.03190.6622 ±0.05090.6632 ±0.0541
ChannelAtt0.8720 ±0.02030.6834 ±0.02780.6107 ±0.03240.6907 ±0.05450.7169 ±0.0624
HybridAtt l0.8885 ±0.01420.7009 ±0.02230.6689 ±0.03140.7264 ±0.04910.7317 ±0.0512
HybridAtt f0.8966 ±0.02140.7082 ±0.02830.6818 ±0.03040.7368 ±0.05910.7424 ±0.0594