Table 1 Test performance of the predictive models.

From: Measuring employment readiness for hard-to-place individuals

 

Admin

ERIQ

Admin+ERIQ

 

AUC-ROC

95% CI

AUC-PR

AUC-ROC

95% CI

AUC-PR

AUC-ROC

95% CI

AUC-PR

Employed within a year

 Linear probability model

62.96

57.33

68.60

8.98

80.69

76.73

84.64

20.02

77.65

73.26

82.04

17.33

 Logistic LASSO

63.63

58.24

69.02

9.31

82.51

78.43

86.59

25.59

82.23

78.27

86.19

21.35

 Random forest

60.60

54.94

66.25

9.01

83.31

79.58

87.05

26.70

84.31

80.85

87.76

24.03

 XGBoost

63.63

57.75

69.50

9.87

83.48

79.72

87.25

27.51

83.73

79.99

87.47

24.89

Outcome rate: 5.51%

Applying for a job

 Linear probability model

67.18

64.38

69.98

41.21

83.00

80.88

85.11

61.04

84.39

82.37

86.41

62.12

 Logistic LASSO

67.67

64.89

70.46

41.61

83.43

81.37

85.49

60.93

85.54

83.62

87.45

63.66

Random forest

69.06

66.31

71.81

43.16

83.90

81.92

85.88

61.16

85.32

83.42

87.21

63.00

XGBoost

68.53

65.76

71.31

42.46

84.24

82.27

86.20

62.76

86.32

84.50

88.13

66.05

Outcome rate: 24.46%

In education within a year

 Linear probability model

80.27

74.30

86.25

20.74

74.99

68.61

81.38

10.16

82.72

76.71

88.73

26.13

 Logistic LASSO

80.74

74.89

86.59

20.06

76.74

70.79

82.70

8.35

86.71

82.20

91.22

28.88

 Random forest

82.67

76.82

88.52

15.74

72.79

65.61

79.98

8.18

85.50

80.38

90.62

19.67

 XGBoost

80.67

74.45

86.88

16.23

75.62

69.24

82.00

9.42

86.63

81.63

91.63

24.87

Outcome rate: 3.07%

Employed or in education within a year

 Linear probability model

68.79

64.15

73.43

19.37

81.05

77.65

84.46

28.22

81.21

77.56

84.86

31.37

 Logistic LASSO

68.62

64.05

73.19

19.17

81.50

78.10

84.89

32.08

82.94

79.47

86.42

36.68

 Random forest

69.17

64.74

73.61

19.96

80.87

77.29

84.44

32.35

85.19

82.26

88.13

38.05

 XGBoost

70.17

65.78

74.57

20.72

81.50

78.04

84.96

33.43

84.06

80.74

87.39

41.26

Outcome rate: 8.42%

  1. The table provides AUC scores for the four models on the two primary and two secondary outcomes.