Table 2 Performance of best models, as they are described by Supplementary Table 8, on various datasets compared to the six-scorer consensus
From: Neural network analysis of sleep stages enables efficient diagnosis of narcolepsy
Test data | Best single model | Mean performance (%) | Best ensemble | Mean performance (%) |
|---|---|---|---|---|
WSC | CC/SH/LS/LSTM/2 | 86.0 ± 5.0 | All CC | 86.4 ± 5.2 |
SSC+KHC, no narcolepsy | CC/LH/SS/LSTM | 76.9 ± 11.1 | All CC | 77.0 ± 11.9 |
SSC+KHC, narcolepsy | CC/LH/SS/LSTM | 68.8 ± 11.0 | All CC | 68.4 ± 12.2 |
IS-RC | CC/LH/LS/LSTM/2 | 84.6 ± 4.6 | All models | 86.8 ± 4.3 |
