Skip to main content

Table 8 Random Forests feature selection aggregate ranking

From: Machine learning can predict survival of patients with heart failure from serum creatinine and ejection fraction alone

Final rank

Feature

Accuracy decrease

Accuracy decrease rank

Gini impurity

Gini impurity rank

1

Serum creatinine

3.78×10−2

1

11.84

1

2

Ejection fraction

3.43×10−2

2

10.71

2

3

Age

1.53×10−2

3

8.58

3

4

Creatinine phosphokinase

7.27×10−4

6

7.26

4

4

Serum sodium

7.20×10−3

4

6.49

6

6

Sex

1.64×10−3

5

1.12

8

6

Platelets

2.47×10−4

8

6.80

5

8

High blood pressure

−1.68×10−3

11

1.13

7

8

Smoking

3.68×10−4

7

0.95

11

10

Anaemia

−5.91×10−4

10

1.06

9

10

Diabetes

−1.41×10−4

9

1.02

10

  1. We merged the two rankings through their position, through the Borda’s method [103]