Table 5 Evaluating performance of different machine learning techniques in disease detection (CAD study)

Bagging

>=0.5

50

4

54

Sensitivity (%)

92.6 (82.4–97.1)

Specificity (%)

81.0 (60.0–92.3)

< 0.5

4

17

21

Positive Predictive Value (PPV) (%)

92.6 (82.4–97.1)

Negative Predictive Value (NPV) (%)

81.0 (60.0–92.3)

Total

54

21

75

Upward Misclassification (%)

19.0 (7.7–40.0)

Downward Misclassification (%)

7.4 (2.9–17.6)

False Omission Rate (FOR) (%)

19.0 (7.7–40.0)

False Discovery Rate (FDR) (%)

7.4 (2.9–17.6)

Positive Likelihood Ratio (LR+)

4.9 (2.0–11.8)

Negative Likelihood Ratio (LR-)

0.1 (0.0–0.2)

Diagnostic Accuracy (%)

89.3 (80.3–94.5)

Disease Prevalence (%)

72.0 (61.0–80.9)

Diagnostic Odds Ratio (DOR)

53.1 (12.0–236.0)

Kappa Agreement

0.7 (0.6–0.9)

Youden’s Index

0.7 (0.6–0.8)

F-score

0.9 (0.8–1.0)

kNN

>=0.5

51

15

66

Sensitivity (%)

94.4 (84.9–98.1)

Specificity (%)

28.6 (13.8–50.0)

< 0.5

3

6

9

Positive Predictive Value (PPV) (%)

77.3 (65.8–85.7)

Negative Predictive Value (NPV) (%)

66.7 (35.4–87.9)

Total

54

21

75

Upward Misclassification (%)

71.4 (50.0–86.2)

Downward Misclassification (%)

5.6 (1.9–15.1)

False Omission Rate (FOR) (%)

33.3 (12.1–64.6)

False Discovery Rate (FDR) (%)

22.7 (14.3–34.2)

Positive Likelihood Ratio (LR+)

1.3 (1.0–1.7)

Negative Likelihood Ratio (LR-)

0.2 (0.1–0.7)

Diagnostic Accuracy (%)

76.0 (65.2–84.2)

Disease Prevalence (%)

72.0 (61.0–80.9)

Diagnostic Odds Ratio (DOR)

6.8 (1.5–30.5)

Kappa Agreement

0.3 (0.0–0.5)

Youden’s Index

0.2 (0.1–0.3)

F-score

0.9 (0.8–0.9)

SVM

>=0.5

52

1

53

Sensitivity (%)

96.3 (87.5–99.0)

Specificity (%)

95.2 (77.3–99.2)

< 0.5

2

20

22

Positive Predictive Value (PPV) (%)

98.1 (90.1–99.7)

Negative Predictive Value (NPV) (%)

90.9 (72.2–97.5)

Total

54

21

75

Upward Misclassification (%)

4.8 (0.8–22.7)

Downward Misclassification (%)

3.7 (1.0–12.5)

False Omission Rate (FOR) (%)

9.1 (2.5–27.8)

False Discovery Rate (FDR) (%)

1.9 (0.3–9.9)

Positive Likelihood Ratio (LR+)

20.2 (3.0–137.0)

Negative Likelihood Ratio (LR-)

0.0 (0.0–0.2)

Diagnostic Accuracy (%)

96.0 (88.9–98.6)

Disease Prevalence (%)

72.0 (61.0–80.9)

Diagnostic Odds Ratio (DOR)

520.0 (44.6–6058)

Kappa Agreement

0.9 (0.8–1.0)

Youden’s Index

0.9 (0.8–1.0)

F-score

1.0 (0.9–1.0)

RF

>=0.5

52

4

56

Sensitivity (%)

96.3 (87.5–99.0)

Specificity (%)

81.0 (60.0–92.3)

< 0.5

2

17

19

Positive Predictive Value (PPV) (%)

92.9 (83.0–97.2)

Negative Predictive Value (NPV) (%)

89.5 (68.6–97.1)

Total

54

21

75

Upward Misclassification (%)

19.0 (7.7–40.0)

Downward Misclassification (%)

3.7 (1.0–12.5)

False Omission Rate (FOR) (%)

10.5 (2.9–31.4)

False Discovery Rate (FDR) (%)

7.1 (2.8–17.0)

Positive Likelihood Ratio (LR+)

5.1 (2.1–12.2)

Negative Likelihood Ratio (LR-)

0.0 (0.0–0.2)

Diagnostic Accuracy (%)

92.0 (83.6–96.3)

Disease Prevalence (%)

72.0 (61.0–80.9)

Diagnostic Odds Ratio (DOR)

110.5 (18.6–657.6)

Kappa Agreement

0.8 (0.6–1.0)

Youden’s Index

0.8 (0.7–0.9)

F-score

0.9 (0.9–1.0)

LR

>=0.5

51

4

55

Sensitivity (%)

94.4 (84.9–98.1)

Specificity (%)

81.0 (60.0–92.3)

< 0.5

3

17

20

Positive Predictive Value (PPV) (%)

92.7 (82.7–97.1)

Negative Predictive Value (NPV) (%)

85.0 (64.0–94.8)

Total

54

21

75

Upward Misclassification (%)

19.0 (7.7–40.0)

Downward Misclassification (%)

5.6 (1.9–15.1)

False Omission Rate (FOR) (%)

15.0 (5.2–36.0)

False Discovery Rate (FDR) (%)

7.3 (2.9–17.3)

Positive Likelihood Ratio (LR+)

5.0 (2.0–12.0)

Negative Likelihood Ratio (LR-)

0.1 (0.0–0.2)

Diagnostic Accuracy (%)

90.7 (82.0–95.4)

Disease Prevalence (%)

72.0 (61.0–80.9)

Diagnostic Odds Ratio (DOR)

72.3 (14.7–355.9)

Kappa Agreement

0.8 (0.6–0.9)

Youden’s Index

0.8 (0.6–0.8)

F-score

0.9 (0.9–1.0)