Skip to main content

Table 5 Evolution of learning from a seed set with 10 rules, based on their F-Measure, AUC-ROC, AU-PRC, and Accuracy

From: Semi-supervised incremental learning with few examples for discovering medical association rules

Iteration

Coincident rules

F-Measure

AUC-ROC

AU-PRC

Accuracy (%)

0

0.55

0.61

0.61

58

1

793

0.70

0.76

0.78

74

2

159

0.71

0.78

0.80

76

3

19

0.74

0.78

0.81

77

4

3

0.73

0.80

0.81

77

5

3

0.71

0.79

0.80

77

6

2

0.72

0.79

0.81

78

7

4

0.75

0.80

0.81

79

  1. Coincident rules are those from the development set that have the same prediction and label based on the p value filter. Best results appear in boldface