A Multilevel Temporal Context Network for Sleep Stage Classification

Table 2

The confusion matrix and per-class result from Sleep-EDF-2013 dataset.

	MLTCN output				Per-class result (%)
	Wake	N 1	N 2	N 3	REM	Pre	Rec	F1
Wake	7427	426	194	18	220	87.4	89.6	88.5
N 1	358	903	748	15	780	50.7	32.2	39.4
N 2	425	220	15531	843	780	88.2	87.3	87.7
N 3	59	1	577	5065	1	85.2	88.8	87.0
REM	226	230	555	4	6702	79.0	86.8	82.7
ACC = 84.2%; MF1 = 77.0%; kappa = 0.78