Skip to main content

Table 3 Best performing models for feature set sizes ranging from 10 to 38.

From: An adaptive genetic algorithm for selection of blood-based biomarkers for prediction of Alzheimer's disease progression

AGA Run Parameters and Result

Random

Stepwise

#

Size

Rho

Subset

Immi- grants

Generations

Feature set**

AUC

AUC

AUC

1

10

0.6

None

30

300

Set_10

0.83

0.59

0.76

2

14

0.1*

Quartile

N/A

300

Set_14

0.85

0.60

0.78

3

18

0.1*

Quartile

N/A

300

Set_18

0.86

0.59

0.79

4

22

0.1*

Quartile

N/A

300

Set_22

0.87

0.59

0.80

5

26

0*

Quartile

N/A

300

Set_26

0.88

0.60

0.81

6

30

0*

Quartile

N/A

300

Set_30

0.89

0.59

0.82

7

34

0.2

None

30

600

Set_34

0.88

0.59

0.82

8

38

0

None

30

300

Set_38

0.85

0.60

0.82

  1. * Penalty applied to the AGA started with only "best quartile" subset.
  2. ** Included features can be found in Additional File 4: Table S2.
  3. The results were compared with models from the randomly selected feature sets (same sizes) and stepwise selected models