A Machine-Learned Predictor of Colonic Polyps Based on Urinary Metabolomics

Table 4

Performance of various prediction algorithms across 5 folds of cross-validation.

Model	Average AUC	AUC standard; deviation
Linear-SVM	0.691	0.016
RBF-SVM	0.690	0.008
Linear-SVM +; Class Cost	0.695	0.014
Naïve Bayes	0.661	0.046
PLS-DA	0.660	0.029
LASSO	0.711	0.013
Random Forest	0.692	0.015
KNN	0.611	0.033
C4.5	0.629	0.027

The bolded row shows the approach we have decided to use.