A time series algorithm to predict surgery in neonatal necrotizing enterocolitis

Table 4 Performance comparison between models with different settings

Loss function		Precision	Recall	F1 score	AP
Cross Entropy		0.872 ± 0.042	0.767 ± 0.060	0.815 ± 0.044	0.864 ± 0.034
Focal Loss	γ = 1	0.952 ± 0.015	0.930 ± 0.015	0.941 ± 0.012	0.978 ± 0.006
	γ = 2	0.949 ± 0.009	0.930 ± 0.026	0.939 ± 0.015	0.967 ± 0.023
	γ = 3	0.942 ± 0.016	0.934 ± 0.023	0.937 ± 0.013	0.972 ± 0.006
	γ = 4	0.937 ± 0.013	0.926 ± 0.015	0.932 ± 0.006	0.974 ± 0.014
	γ = 5	0.941 ± 0.011	0.922 ± 0.025	0.931 ± 0.013	0.974 ± 0.008
	γ = 6	0.942 ± 0.020	0.926 ± 0.023	0.933 ± 0.007	0.974 ± 0.009
	γ = 7	0.935 ± 0.037	0.871 ± 0.024	0.902 ± 0.020	0.953 ± 0.016
	γ = 8	0.884 ± 0.069	0.750 ± 0.051	0.809 ± 0.034	0.857 ± 0.040

ISSN: 1472-6947