Table 2 Feature’s ranked based on five different approaches.

From: A comparison of machine learning algorithms and traditional regression-based statistical modeling for predicting hypertension incidence in a Canadian population

Feature

Ranking based on random survival forest relative importance

Ranking based on statistical equivalent signature

Ranking based on Harrel’s C-index/Somers’ Dxy rank correlation

Ranking based on Lasso Cox coefficients/variable importance

Ranking based on univariate Cox p values

Systolic blood pressure

1

1

1

13

1

Diastolic blood pressure

2

20

2

15

5

Body mass index

3

2

3

11

3

Waist-hip ratio

4

11

5

1

4

Diabetes

5

5

14

3

10

Cardiovascular disease

6

3

16

2

9

Age

7

4

4

14

2

Job schedule

8

6

6

4

7

Working Status

9

8

7

19

8

Total household income

10

7

9

6

6

Residence

11

13

10

5

12

Total sleep time

12

9

11

22

15

Highest education level completed

13

12

8

10

11

Family history of hypertension

14

17

18

12

16

Physical activity, quartiles

15

19

22

21

23

Smoking status

16

14

12

23

14

Total physical activity time

17

24

15

16

17

Depression,

18

21

21

9

24

Ethnicity

19

10

24

18

21

Sex

20

18

13

8

13

Total sitting time

21

22

23

17

22

Alcohol consumption

22

16

17

7

19

Marital status

23

15

20

24

20

Vegetable and fruit consumption

24

23

19

20

18