A Wavelet-Based Learning Model Enhances Molecular Prognosis in Pancreatic Adenocarcinoma

Table 3

Performance comparison among the eight typical wavelet-based and classic LASSO and SWT-CNN methods.

Method	Basis	AUC at 3 years	AUC at 5 years	Predictors
haar	haar	0.618	0.756	5
dmeyer	dmeyer	0.823	0.869	7
symlets	sym3	0.806	0.904	5
coiflets	coif4	0.696	0.788	5
daubechies	db7	0.804	0.867	5
biorSplines	bior1.5	0.787	0.937	5
reverseBior	rbior2.8	0.852	0.897	10
Fejer-Korovkin	fk14	0.672	0.861	5
SWT-CNN	sym4	0.721	0.859	8
LASSO		0.782	0.802	11