Table 7 Second layer model (LightGBM and other models) test results for different diseases (threshold: 0.5).

From: Emerging infectious disease surveillance using a hierarchical diagnosis model and the Knox algorithm

Category of infectious diseases

The number of infectious diseases predicted by the first-level model(TP1K + FP1k)

Sensitivity2k

FNR2k

FPR2k

LGBM (%)

XGBoost (%)

RF (%)

LGBM (%)

XGBoost (%)

RF (%)

LGBM (%)

XGBoost (%)

RF (%)

Viral hepatitis

16,207

90.59

95.73

89.04

9.41

4.27

10.96

61.00

98.00

74.00

Influenza

3008

96.99

89.40

84.87

3.01

10.60

15.13

36.73

81.00

89.00

Tuberculosis

4101

92.35

69.09

92.93

7.65

30.91

7.07

26.00

67.00

94.00

Hand-foot-and-mouth disease

1385

75.20

61.68

58.31

24.80

38.32

41.69

26.00

71.00

87.00

Syphilis

855

95.62

90.69

94.24

4.38

9.31

5.76

50.00

86.00

92.00

Infectious diarrhea

525

77.22

61.31

62.06

22.78

38.69

37.94

35.00

93.00

95.00

Scarlatina fever

410

74.14

40.39

69.06

25.86

59.61

30.94

69.00

93.00

94.00

COVID-19

80

93.33

96.00

92.00

6.67

4.00

8.00

54.55

100.00

95.45

Weighted average by test sample size

90.44

87.43

86.90

9.56

12.57

13.10

50.26

89.33

80.83