Table 2 Toxicity dataset, mean accuracy of 10-Fold CV with 100 replications for feature sets with cardinality ranging from 2 to 20.

From: Structure-based design and classifications of small molecules regulating the circadian rhythm period

Features

10 Fold CV—accuracy (%)

DTC

RFC

ETC

XGBC

2

60.22

68.63

70.32

65.06

3

59.20

68.68

70.40

67.64

4

60.39

71.97

68.82

68.10

5

71.93

69.47

69.88

68.88

6

69.87

71.18

67.99

68.67

7

71.73

72.70

68.12

68.20

8

72.81

72.78

69.16

67.37

9

75.87

72.99

68.16

68.32

10

75.65

72.68

70.21

67.92

11

76.75

71.71

68.60

70.18

12

75.33

71.43

68.83

70.16

13

76.46

72.36

68.57

68.87

14

78.49

72.54

68.23

68.63

15

77.22

70.76

69.67

69.08

16

77.21

72.41

68.73

71.16

17

75.83

72.81

68.77

71.17

18

78.75

72.37

69.49

71.05

19

78.77

70.03

70.32

70.11

20

78.02

70.73

71.36

70.73

  1. DTC, RFC, ETC, and XGBC trained and tested on feature sets with cardinality between 2 and 20.