Journal of Cheminformatics

Table 2 Prediction of the DFT DM by random forests on the basis of different molecular descriptors

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

Descriptors (#)	Training set^a		Test set
Descriptors (#)	MAE (D)	R²/RMSE (D)	MAE (D)	R²/RMSE (D)
RDF_N^b (384)	0.944	0.480/1.332	0.947	0.498/1.344
RDF_P^c (384)	0.890	0.512/1.295	0.882	0.549/1.287
PchmDM_N^b (360)	0.924	0.545/1.267	0.880	0.589/1.250
PchmDM_P^c (360)	0.873	0.569/1.240	0.931	0.566/1.278
CDK^d (47)	0.983	0.434/1.385	0.985	0.445/1.402
MACCS FP^e (166)	0.790	0.579/1.195	0.775	0.609/1.182
PubChem FP^e (881)	0.817	0.547/1.238	0.801	0.584/1.217
CDK FP^e (1024)	0.880	0.501/1.301	0.874	0.521/1.305

^aOOB estimation
^bDescriptors calculated using NBO charges
^cDescriptors calculated using PEOE charges
^dGeometric CDK descriptors
^eFingerprints

Back to article page

ISSN: 1758-2946

Contact us

Submission enquiries: journalsubmissions@springernature.com