Skip to main content
. 2020 Jun 15;8(3):215–250. doi: 10.5599/admet.835

Table 1.

Statistical evaluations of ML predictions for the 9 individual predictors and for four possible definitions of the Vox Machinarum classifier over our internal validation set of 36 compounds

Method RMSE AAE R 2 Err < 0.5 Err < 1.0
Extra Trees 0.766 0.582 0.789 18 (50%) 30 (83%)
Random Forest 0.766 0.607 0.792 17 (47%) 29 (81%)
Bagging 0.827 0.659 0.737 16 (44%) 26 (72%)
MLP 1.017 0.804 0.597 16 (44%) 24 (67%)
kNN man unw k4 1.034 0.792 0.579 18 (50%) 26 (72%)
kNN eu exp k6 1.054 0.818 0.573 14 (39%) 25 (69%)
kNN man exp k4 1.062 0.791 0.548 18 (50%) 25 (69%)
RVM 1.121 0.820 0.509 16 (44%) 25 (69%)
kNN eu unw k6 1.122 0.906 0.528 12 (33%) 22 (61%)
Vox Machinarum (3) 0.760 0.602 0.797 17 (47%) 29 (81%)
Vox Machinarum (5) 0.787 0.627 0.771 17 (47%) 29 (81%)
Vox Machinarum (7) 0.891 0.695 0.695 16 (44%) 29 (81%)
Vox Machinarum (9) 0.944 0.728 0.660 16 (44%) 28 (78%)