Table 4.

Sleep/wake differentiation performance by random forest classifier across different feature inputs in the Apple Watch (PPG, MEMS) dataset

	Accuracy	Wake correct (specificity)	Sleep correct (sensitivity)	κ	AUC
Motion	0.793	0.713	0.8	0.27	0.81
	0.869	0.53	0.9	0.329
	0.891	0.457	0.93	0.346
	0.904	0.399	0.95	0.352
HR	0.771	0.454	0.8	0.142	0.708
	0.849	0.282	0.9	0.152
	0.872	0.221	0.93	0.149
	0.886	0.174	0.95	0.14
Motion, HR	0.792	0.707	0.8	0.267	0.816
	0.869	0.519	0.9	0.322
	0.89	0.448	0.93	0.339
	0.904	0.394	0.95	0.349
Motion, HR, and Clock Proxy	0.799	0.789	0.8	0.303	0.871
	0.879	0.653	0.9	0.405
	0.901	0.579	0.93	0.433
	0.914	0.513	0.95	0.444

Fraction of wake correct, fraction of sleep correct, accuracy, κ, and AUC for sleep-wake predictions of random forest classifier with use of motion, HR, clock proxy, or combination of features. HR, heart rate.