Skip to main content

Table 5 Top 10 important descriptors selected by three blending methods

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

Uniform blending

Linear blending

Any blending

Selected descriptors

Feature importance

Selected descriptors

Feature importance

Selected descriptors

Feature importance

Gap

0.1654

Gap

0.1330

Gap

0.1572

AP(xx)

0.0470

AP(xx)

0.0359

AP(xx)

0.0411

F01[C-N]

0.0216

F01[C-N]

0.0247

Solvent

0.0224

SpMax2_Bh(i)

0.0176

SpMax2_Bh(i)

0.0174

F01[C-N]

0.0198

Solvent

0.0169

Solvent

0.0141

SpMax2_Bh(i)

0.0170

P_VSA_MR_7

0.0162

F02[C-N]

0.0133

P_VSA_ppp_L

0.0147

P_VSA_ppp_L

0.0151

EP(xx)

0.0122

EP(xx)

0.0115

EP(xx)

0.0133

F01[C-C]

0.0111

F02[C-N]

0.0108

Chi1_EA(dm)

0.0119

SpMin5_Bh(m)

0.0106

Chi1_EA(dm)

0.0105

F02[C-N]

0.0109

P_VSA_ppp_L

0.0103

Chi0_EA(dm)

0.0092

  1. Italic values indicate the significance of important features which highly affect the fluorescence wavelengths