Fig. 3
From: TB-IECS: an accurate machine learning-based scoring function for virtual screening
The performance of models trained on theory-based features and on singe SFs and tested on Dataset I.(1) All: the model trained on all the energy components of the 15 SFs with feature selection, (2) Formula: the best model among the models using Formula-based feature combinations, (3) Tree-mean and Tree-sum: the best model among the models using Tree-based feature combinations, (4) Tree-*-rank: models trained on the features ranked first according to the importance, (5) others: models built on single SFs’ energy components