Table 2 Ranked feature importance for the XGBoost, gradient boosting (GB), random forest (RF), and deep neural network (DNN) models.

Criminal conviction of either parent

18

20

19

20

19.3

Sex

15

19

18

21

18.3

ADHD relative

21

15

15

18

17.3

Number of academic subjects failed

9

21

20

19

17.3

Speech/learning disability

20

18

17

10

16.3

Autism disorder

19

16

16

12

15.8

Depression

16

14

13

16

14.8

Depression relative

13

17

14

15

14.8

Head circumference

3

13

21

13

12.5

Alcohol disorder relative

14

12

12

6

11.0

Anxiety

11

11

11

9

10.5

Criminal conviction

10

8

6

17

10.3

Motor/tic disorders

17

10

8

0

8.8

Allergic rhinitis and Allergic conjunctivitis

7

6

5

14

8.0

Asthma relative

8

9

10

4

7.8

Sleep disorders

12

7

4

8

7.8

Anxiety relative

2

5

9

11

6.8

Allergic dermatitis

6

2

3

7

4.5

Substance use disorders relative

0

3

7

5

3.8

Eating disorders

5

4

1

3

3.3

Small size for age

4

1

2

2

2.3

Eating disorders relative

1

0

0

1

0.5

Kendall’s τb

XGB

GB

RF

DNN

Average

XGB

1

-

-

-

-

GB

0.56^*** (0.32–0.76)

1

-

-

-

RF

0.42^** (0.11–0.70)

0.81^*** (0.64–0.95)

1

-

-

DNN

0.32^* (0.02–0.58)

0.55^*** (0.30–0.75)

0.52^*** (0.27–0.72)

1

-

Average

0.62^*** (0.39–0.81)

0.88^*** (0.77–0.96)

0.78^*** (0.60–0.92)

0.65^*** (0.42–0.85)

1