Skip to main content

Table 8 Performance comparison of ML models by feature sets in predicting mortality from COVID-19

From: The predictive power of data: machine learning analysis for Covid-19 mortality based on personal, clinical, preclinical, and laboratory variables in a case–control study

Feature set

Model

Sensitivity

Specificity

Accuracy

F1-score

AUC

Demographic

DT

63.26 (13.59)

66.48 (12.19)

65.02 (2.60)

63.26 (13.59)

64.87 (2.24)

XGBoost

60.50 (3.79)

68.30 (3.95)

64.83 (2.58)

60.50 (3.79)

64.40 (2.57)

SVM

66.12 (3.15)

63.08 (3.59)

64.36 (1.67)

66.12 (3.15)

64.60 (1.48)

NB

62.48 (3.77)

63.97 (5.08)

63.29 (2.33)

62.48 (3.77)

63.23 (2.10)

RF

61.37 (2.57)

70.58 (3.96)

66.49 (1.84)

61.37 (2.57)

65.97 (1.77)

Clinical & Conditions

DT

90.85 (4.22)

86.04 (5.03)

88.24 (1.58)

90.85 (4.22)

88.45 (1.47)

XGBoost

92.74 (2.46)

92.96 (2.37)

92.82 (0.77)

92.74 (2.46)

92.85 (0.70)

SVM

90.70 (3.24)

92.61 (2.12)

91.72 (1.03)

90.70 (3.24)

91.66 (1.14)

NB

88.40 (2.58)

94.32 (1.87)

91.63 (0.75)

88.40 (2.58)

91.36 (0.85)

RF

92.88 (2.51)

92.85 (2.08)

92.82 (0.87)

92.88 (2.51)

92.86 (0.87)

Comorbidities

DT

74.83 (1.88)

79.19 (1.80)

77.35 (0.47)

74.83 (1.88)

77.01 (0.32)

XGBoost

77.45 (1.32)

83.43 (1.28)

80.88 (0.57)

77.45 (1.32)

80.44 (0.74)

SVM

74.83 (1.88)

78.93 (1.38)

77.19 (0.39)

74.83 (1.88)

76.88 (0.46)

NB

75.58 (1.89)

79.76 (1.18)

77.98 (0.16)

78.58 (1.89)

77.67 (0.44)

RF

75.19 (1.50)

81.96 (1.84)

79.08 (1.04)

75.19 (1.50)

78.58 (1.20)

Treatment

DT

75.17 (4.35)

87.62 (2.60)

81.94 (1.81)

75.17 (4.35)

81.39 (1.90)

XGBoost

78.42 (3.43)

89.27 (1.58)

84.29 (1.59)

78.42 (3.43)

83.84 (1.69)

SVM

72.89 (2.39)

91.15 (1.34)

82.82 (1.30)

72.89 (2.39)

82.02 (1.36)

NB

72.33 (3.64)

88.50 (2.25)

81.13 (2.04)

72.33 (3.64)

80.42 (2.10)

RF

79.17 (3.32)

89.55 (1.36)

84.80 (1.52)

79.17 (3.32)

84.36 (1.61)

Initial vital signs

DT

90.30 (2.52)

98.53 (1.91)

95.77 (1.11)

92.30 (2.52)

95.42 (1.17)

XGBoost

95.85 (1.97)

99.83 (0.53)

98.06 (0.97)

95.85 (1.97)

97.84 (1.06)

SVM

94.45 (1.61)

99.49 (0.62)

97.24 (1.02)

94.45 (1.61)

96.97 (1.06)

NB

87.37 (2.09)

99.21 (0.76)

93.95 (1.19)

87.37 (2.09)

93.29 (1.24)

RF

94.63 (2.02)

99.83 (0.54)

97.52 (1.12)

94.63 (2.02)

97.23 (1.20)

Symptoms

DT

92.79 (3.69)

97.09 (1.39)

95.24 (2.05)

92.79 (3.69)

94.94 (2.25)

XGBoost

97.08 (1.32)

98.76 (0.79)

98.03 (0.78)

97.08 (1.32)

97.92 (0.78)

SVM

91.78 (2.48)

98.02 (1.05)

95.27 (1.38)

91.78 (2.48)

94.90 (1.45)

NB

82.03 (4.83)

90.19 (3.29)

86.58 (2.19)

82.03 (4.83)

86.11 (2.32)

RF

95.55 (2.17)

97.82 (0.51)

96.83 (1.01)

95.55 (2.17)

96.69 (1.14)

Laboratory test

DT

100 (0.0)

100 (0.0)

100 (0.0)

100 (0.0)

100 (0.0)

XGBoost

100 (0.0)

100 (0.0)

100 (0.0)

100 (0.0)

100 (0.0)

SVM

99.80 (0.33)

100 (0.0)

99.91 (0.15)

99.80 (0.33)

99.90 (0.16)

NB

100 (0.0)

100 (0.0)

100 (0.0)

100 (0.0)

100 (0.0)

RF

100 (0.0)

100 (0.0)

100 (0.0)

100 (0.0)

100 (0.0)

  1. The average values are expressed from the test set as the Mean (SD)
  2. DT Decision Tree, XGBoost eXtreme Gradient Boosting, SVM Support Vector Machine, NB Naïve Bayes, RF Random Forest