Skip to main content

Table 3 Comparison of kNN classification model and SVM models

From: Open-source QSAR models for pKa prediction using multiple machine learning approaches

Model algorithm

Descriptor type

Variables

Train BA

Fivefold CV BA

Test BA

kNN

Continuous

15

0.8

0.8

0.77

SVM

Continuous

15

0.92

0.8

0.73

SVM

Continuous

511

0.98

0.79

0.72

SVM

Fingerprints

1565

0.98

0.8

0.74

SVM

Fragment counts

815

0.96

0.8

0.73

  1. The kNN classification model used the 15 GA selected descriptors. The SVM models used the same descriptors as well as additional continuous descriptors, fingerprints, and fragment counts as noted in the table
Back to article page

Journal of Cheminformatics

ISSN: 1758-2946

Contact us