Table 7.

Baseline model performance on time-domain and frequency-domain inputs for ICU vs FU classification.

Variation	FE domain	Model name	Accuracy	Macro precision	Macro recall	Macro F1	Weighted F1
10 wavelength Segment	Time-domain	DT	0.614	0.611	0.609	0.609	0.613
		RF (10)	0.654	0.655	0.636	0.632	0.642
		RF (100)	0.730	0.732	0.721	0.722	0.722
		SVM (RBF)	0.675	0.673	0.663	0.664	0.670
		SVM (poly)	0.594	0.637	0.553	0.495	0.518
		MLP (10)	0.593	0.585	0.553	0.582	0.587
		MLP (100)	0.620	0.616	0.612	0.612	0.618
		MLP (500)	0.628	0.627	0.624	0.622	0.626
5 wavelength Segment (truncated)	Time-domain	DT	0.633	0.630	0.630	0.629	0.633
		RF (10)	0.677	0.681	0.664	0.662	0.669
		RF (100)	0.753	0.754	0.746	0.747	0.751
		SVM (RBF)	0.719	0.720	0.707	0.708	0.715
		SVM (poly)	0.650	0.677	0.618	0.601	0.617
		MLP (10)	0.642	0.638	0.636	0.635	0.640
		MLP (100)	0.666	0.664	0.661	0.661	0.665
		MLP (500)	0.665	0.662	0.661	0.660	0.664
Best performing Model frequency bins	Frequency-domain	RF (100)–4	0.692	0.689	0.685	0.685	0.690
		RF (100)–8	0.697	0.696	0.692	0.692	0.696
		RF (100)–16	0.849	0.848	0.847	0.847	0.849
		RF (100)–32	0.863	0.862	0.861	0.861	0.862
		RF (100)–64	0.875	0.875	0.874	0.874	0.875
		RF (100)–128	0.882	0.881	0.881	0.880	0.882
		RF (100)–256	0.882	0.882	0.880	0.880	0.882
		RF (100)–512	0.858	0.859	0.856	0.856	0.858
		RF (100)–1024	0.835	0.837	0.831	0.832	0.835