Skip to main content

Table 5 Results of data set II.

From: Drug activity prediction using multiple-instance learning via joint instance and feature selection

Iteration

0

1

2

3

4

5

Method 1

Error rate (%)

CV

12.50

10.00

10.00

11.10

8.63

-

  

Training

6.73

6.23

9.23

8.73

7.48

-

  

Test

16.18

15.44

14.71

14.71

15.44

-

 

Number of instances

35434

58

40

26

23

-

 

Number of features

14002

106

43

35

19

-

Method 2

Error rate (%)

CV

12.50

8.75

10.00

10.00

12.50

12.50

  

Training

4.99

6.98

11.97

11.47

13.72

13.72

  

Test

16.91

19.85

14.71

15.81

16.18

16.18

 

Number of instances

35434

62

37

10

7

1

 

Number of features

14002

92

36

20

10

6

  1. Note: Iteration 0 is the original MILES method without instance and feature selection.