Journal of Cheminformatics

Table 6 RF prediction of DM with subsets of descriptors from models C and F

From: Machine learning for the prediction of molecular dipole moments obtained by density functional theory

Model/no descriptors	MAE (D)	R²/RMSE (D)
Training set
C/75^a	0.515	0.826/0.769
C/100^a	0.517	0.826/0.771
C/125^a	0.518	0.826/0.7709
C/32^b	0.521	0.824/0.773
C/39^c	0.523	0.824/0.775
C/296^d	0.529	0.821/0.782
F/75^a	0.482	0.844/0.731
F/100^a	0.482	0.844/0.731
F/125^a	0.483	0.844/0.732
F/34^b	0.501	0.838/0.745
F/41^c	0.499	0.838/0.743
F/297^d	0.503	0.832/0.755
Test set
C/75^a	0.502	0.847/0.744
C/100^a	0.506	0.846/0.748
C/125^a	0.505	0.845/0.748
C/32^b	0.509	0.845/0.747
C/39^c	0.512	0.844/0.749
C/296^d	0.519	0.843/0.758
F/75^a	0.468	0.864/0.704
F/100^a	0.466	0.865/0.702
F/125^a	0.466	0.867/0.699
F/34^b	0.481	0.860/0.713
F/41^c	0.479	0.860/0.710
F/297^d	0.485	0.852/0.725

^aOOB estimation for the training set
^bUsing the mean decrease in accuracy measure of importance for the descriptors in the RF algorithm
^cUsing the the CFS with BestFirst routine from Weka
^dUsing the the CFS with GreedyStepwise routine from Weka
^eUsing the the CFS with PSOSearch routine from Weka

Back to article page

ISSN: 1758-2946

Contact us

Submission enquiries: journalsubmissions@springernature.com