Table 3 Ranked feature importance for the deep neural network (DNN) model.

From: Predicting childhood and adolescent attention-deficit/hyperactivity disorder onset: a nationwide deep learning approach

Rank

Males

Females

21

Criminal conviction of either parent

Criminal conviction of either parent

20

Number of academic subjects failed

Number of academic subjects failed

19

ADHD relative

Head circumference

18

Depression relative

ADHD relative

17

Depression

Depression relative

16

Head circumference

Allergic rhinitis and allergic conjunctivitis

15

Allergic rhinitis and allergic conjunctivitis

Criminal conviction

14

Anxiety

Anxiety relative

13

Anxiety relative

Anxiety

12

Autism disorder

Autism disorder

11

Criminal conviction

Depression

10

Allergic dermatitis

Sleep disorders

9

Sleep disorders

Allergic dermatitis

8

Eating disorders

Alcohol disorder relative

7

Speech/learning disability

Asthma relative

6

Alcohol disorder relative

Speech/learning disability

5

Substance use disorders relative

Substance use disorders relative

4

Asthma relative

Motor/tic disorders

3

Motor/tic disorders

Small size for age

2

Small size for age

Eating disorders relative

1

Eating disorders relative

Eating disorders

  1. Rank based on SHAP feature importance (mean absolute Shapley values). Importance ranging from 1 (less important) to 21 (most important).