Research Article

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

Table 3

Performance of LASSO Classifier using various normalization and transformation methods.

MethodAverage AUCAUC standard
deviation

None0.6800.009
Log0.7110.013
Creatinine + Log0.6980.019
Sum + Log0.7010.018
Vector Length + Log0.7030.014
PQ + Log0.6700.024

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