Skip to main content

Table 4 We report the performance of four classifiers in one experiment with best values per row shown in bold font

From: A new concordant partial AUC and partial c statistic for imbalanced data in the evaluation of machine learning algorithms

Measures LDA LogR SVM NN NN-SVM
Whole Area
AUC 82.9% 77.1% 84.8% 86.0% 1.2%
AUPRC+ 60.9% 53.5% 72.2% 71.0% −1.2%
AUPRC− 54.5% 56.7% 53.7% 53.3% −0.4%
Partial Area i = 1
sPA 75.0% 69.2% 78.8% 79.2% 0.4%
pAUC 19.2% 16.0% 21.3% 21.6% 0.3%
pAUCc 47.5% 37.2% 49.5% 48.0% 1.5%
Partial Area i = 2
sPA 90.0% 82.2% 89.4% 92.2% 2.8%
pAUC 29.7% 27.1% 29.5% 30.4% 0.9%
pAUCc 18.5% 22.9% 17.4% 21.0% 3.6%
Partial Area i = 3
sPA 100% 100% 99.7% 100% 0.3%
pAUC 34.0% 34.0% 34.0% 34.0% 0%
pAUCc 17.0% 17.0% 17.9% 17.0% 0.9%
sPA: sum of NN-SVM 3.5%
pAUC: sum of NN-SVM 1.2%
pAUCc: sum of NN-SVM 1.2%