Skip to main content

Table 3 Top 10 important descriptors selected by four DT-based ensemble learning models

From: Comparison and improvement of the predictability and interpretability with ensemble learning models in QSPR applications

RFExtraTrees
Selected descriptorsFeature importanceSelected descriptorsFeature importance
Gap0.3412Gap0.0712
AP(xx)0.0986F01[C-N]0.0350
Chi1_EA(dm)0.0344AP(xx)0.0243
Chi0_EA(dm)0.0274SpMax2_Bh(i)0.0221
EP(xx)0.0239F02[C-N]0.0192
P_VSA_ppp_L0.0215SpMax7_Bh(m)0.0179
SpDiam_AEA(ed)0.0160F01[C-C]0.0174
SpMax_AEA(ed)0.0134C-0040.0157
SpMin5_Bh(m)0.0119P_VSA_e_20.0152
CATS2D_06_LL0.0093EP(xx)0.0123
AdaBoostGBM
Selected descriptorsFeature importanceSelected descriptorsFeature importance
Gap0.1196Gap0.1621
AP(xx)0.0682Solvent0.0534
P_VSA_MR_70.0601MATS1e0.0147
SpMax2_Bh(i)0.0408Chi1_EA(dm)0.0108
F01[C-N]0.0382MATS6m0.0107
P_VSA_ppp_L0.0317SpMax_AEA(ed)0.0102
F02[C-N]0.0224Eig01_AEA(ed)0.0100
LUMO0.0197CATS2D_00_LL0.0092
EP(xx)0.0167AP(xx)0.0089
SdsCH0.0141SpMin8_Bh(e)0.0086