Table 5 Information-gain-based dimension reduction in the three machine learning models.

From: Performance and efficiency of machine learning algorithms for analyzing rectangular biomedical data

Accuracy

Alive (%)

Non-breast cancer (%)

Breast cancer (%)

CVS (%)

Other cause (%)

Total accuracy (%)

Time, mean (s)

Feature change

Random forests

 Raw data (54 features)

96.69

0.00

43.22

2.25

1.09

70.23

55.57

 Delete IG < 0.0002(46 features)

96.66

0.00

42.14

2.25

0.54

69.99

46.63

 Delete IG < 0.0005(40 features)

96.52

0.00

42.14

1.93

0.27

69.85

37.03

 Delete IG < 0.001(32 features)

96.52

0.00

41.87

1.93

0.27

69.81

25.99

 Delete IG < 0.005(21 features)

96.22

0.00

41.19

2.89

0.27

69.57

14.59

 Delete IG < 0.01(15 features)

96.59

0.00

40.11

2.57

1.09

69.68

10.15

 Delete IG < 0.02(10 features)

97.89

0.00

33.33

0.64

1.09

69.28

5.94

 Delete top three (7 features)

97.75

0.00

31.84

0.00

0.00

68.81

3.85

Artificial neural networks

 Raw data (54 features)

95.47

0.00

44.04

5.79

5.45

70.16

12.92

 Delete IG < 0.0002(46 features)

95.29

0.00

43.36

0.64

4.09

69.48

8.92

 Delete IG < 0.0005(40 features)

95.36

0.00

44.17

4.50

7.36

70.19

10.32

 Delete IG < 0.001(32 features)

95.74

0.00

42.41

4.50

5.18

69.96

10.48

 Delete IG < 0.005(21 features)

95.77

0.00

41.60

3.54

2.18

69.54

12.15

 Delete IG < 0.01(15 features)

96.73

0.00

39.70

2.57

0.54

69.65

8.83

 Delete IG < 0.02(10 features)

97.82

0.00

33.88

0.00

0.82

69.25

9.66

 Delete top three (7 features)

97.92

0.00

29.95

0.00

0.00

68.61

11.30

Support vector machine

 Raw data (54 features)

96.22

0.00

39.43

0.00

0.00

69.06

175.50

 Delete IG < 0.0002(46 features)

96.22

0.00

39.43

0.00

0.00

69.06

135.55

 Delete IG < 0.0005(40 features)

96.18

0.00

39.43

0.00

0.27

69.06

100.38

 Delete IG < 0.001(32 features)

96.22

0.00

39.43

0.00

0.27

69.08

67.81

 Delete IG < 0.005(21 features)

96.28

0.00

39.02

0.00

0.00

69.03

45.18

 Delete IG < 0.01(15 features)

96.15

0.00

38.75

0.00

2.18

69.08

32.37

 Delete IG < 0.02(10 features)

97.68

0.00

31.71

0.00

0.00

68.74

28.04

 Delete top three (7 features)

97.89

0.00

31.44

0.00

0.00

68.83

18.26

  1. IG information gain.