Table 4.

Performance comparison of different feature selection methods and machine learning models for GTR patients.

Model	Accuracy	p (Binomial)	MSE	Median err.	SpearmanR
VIF-BASED FEATURE SUBSET
Regression	0.47	0.01	28154236838	1,112	0.18
Regr. BH	0.46	0.07	109618	148	0.46
Lasso	0.36	0.60	2293591760	557	0.05
Ridge	0.33	1.0	16655918658	672	−0.02
kNN	0.30	0.53	159553	223	−0.08
RFR	0.35	0.75	149299	207	0.15
SVR	0.27	0.20	140181	189	−0.77
SVC	0.40	0.17	194331	445	0.06
FEATURES EXTRACTED by PCA
Regression	0.39	0.17	672193	478	0.03
Lasso	0.36	0.59	688037	559	0.04
Ridge	0.38	0.25	558488	457	0.02
kNN	0.34	0.92	149014	194	0.07
RFR	0.34	0.92	163826	218	0.02
SVR	0.27	0.20	140298	189	−0.79
SVC	0.40	0.17	198655	445	0.05

Regr. BH, Regression on all features that were significant after Benjamini–Hochberg multiple test correction.