Table 2 Rank of top 11 important variables selected from various machine learning methods; support vector machine with recursive feature elimination (SVM-RFE), logistic regression with recursive feature elimination (LR-RFE), random forest using gini index (RF-gini), and random forest using information entropy (RF-entropy).

From: Development of early prediction model for pregnancy-associated hypertension with graph-based semi-supervised learning

Clinical variables

Ranks by selection methods

Integrated rankinga

SVM (RFE)

LR (RFE)

RF (gini)

RF (entropy)

Diastolic blood pressure (BP) in early pregnancy

1

1

1

1

1.00

Systolic BP in early pregnancy

1

2

2

2

1.68

Diastolic BP in late first trimester

3

5

4

4

3.94

Hemoglobin level measured in the first trimester

4

14

5

5

6.17

Systolic BP in late first trimester

16

17

3

2

6.36

BMI before pregnancy

7

3

10

10

6.77

Maternal age

5

4

14

14

7.91

BMI in late first trimester

10

9

9

6

8.35

History of preeclampsia in previous pregnancy

12

6

6

12

8.49

Weight in late first trimester

8

7

12

11

9.27

Weight before pregnancy

6

10

11

13

9.62

  1. Early pregnancy, measured at 7.7 ± 1.2 weeks; late first trimester, measured at 12.4 ± 0.5 weeks.
  2. BMI body mass index, BP blood pressure.
  3. aTo combine/aggregate four different rankings, we apply the geometric mean which is defined as \({\left({\prod }_{i-1}^{n}{r}_{i}\right)}^\frac{1}{n}=\sqrt[n]{{r}_{1}{r}_{2}\dots {r}_{n}}\) where \({r}_{i}\) is the variable ranks in \(i\)th selection methods.
Back to article page