Skip to main content

Table 4 Five-fold cross-validated prediction accuracies for the CIP and CFP descriptors on entire dataset (2232 compounds)

From: Predicting anatomic therapeutic chemical classification codes using tiered learning

Fivefold cross validation results across entire dataset

ATC Level

1

2

3

4

Descriptor

CIP

CFP

CIP

CFP

CIP

CFP

CIP

CFP

MLP

65.82%

32.38%

58.59%

31.06%

53.02%

27.79%

39.58%

24.07%

Naive Bayes

46.48%

22.51%

32.23%

17.41%

30.13%

14.81%

24.30%

8.91%

Random Forest

60.08%

37.51%

53.45%

28.77%

47.01%

25.28%

36.11%

22.68%

SVM

64.69%

39.04%

56.53%

31.92%

52.27%

28.69%

39.28%

25.60%

  1. At each level, the highest prediction accuracy is highlighted in bold