Table 2 Mean Absolute Error (MAE) – calculated by 10-fold cross-validation – in the prediction of log Vss by each regression method, in model selection dataset (with 402 compounds)
Regression method | MAE (logVss) | ||
|---|---|---|---|
Descriptor set includes predicted log K t:p ’s | Descriptor set without predicted log K t:p ’s | Descriptor set selected by genetic CFS | |
M5P-4 | 0.3891 | 0.3698 | 0.3739 |
M5P-6 | 0.3823 | 0.3665 | 0.4003 |
M5P-8 | 0.4715 | 0.3616 | 0.3847 |
M5P-10 | 0.4751 | 0.3658 | 0.3763 |
M5P-RegTree | 0.3689 | 0.3772 | 0.3782 |
REPTree | 0.3824 | 0.4330 | 0.3974 |
M5Rules | 0.3911 | 0.3843 | 0.3836 |
Bagging (M5P) | 0.3713 | 0.3371 | 0.3509 |