Skip to main content
. 2016 Feb 9;6(2):e732. doi: 10.1038/tp.2015.221

Figure 1.

Figure 1

Forward feature selection results with and without parameter tuning. Using forward feature selection methods, we tested the performance of six machine learning algorithms at each feature addition, both with parameter tuning (a) and without (b). All of the six algorithms performed with peak AUC⩾0.93; five of these only required five features. AUC, area under the receiver operating characteristic curve; LDA, linear discriminant analysis; SVC, support vector classification.