Table 3.

Confusion matrix and overall performance on the sleep-edf dataset.

Stage	Predictions					Per-Class Metrics
	Wake	N1	N2	N3	REM	Precision	Recall	F1
W	2388	33	6	1	5	99.1	98.2	98.6
N1	15	83	25	1	35	52.9	52.2	52.5
N2	2	28	1024	49	16	92.7	91.5	92.1
N3	2	0	44	336	1	86.8	87.7	87.3
REM	2	13	6	0	437	88.5	95.4	91.8

Overall accuracy: 93.7%, MF1 score: 84.5.