Skip to main content

Table 1 Combinations of γ, p, p(rel) and β used for the simulation design

From: Cost-Constrained feature selection in binary classification: adaptations for greedy forward selection and genetic algorithms

 

γ

p

p(rel)

β

Setting A

\(\frac {1}{2}\)

30

18

0.3

Setting B

\(\frac {2}{3}\)

30

3

1

Setting C

\(\frac {1}{3}\)

300

30

0.5

Setting D

\(\frac {2}{3}\)

300

3

0.5

Setting E

2

1500

15

0.5

Setting F

\(\frac {1}{2}\)

1500

20

0.5

Setting G

\(\frac {1}{3}\)

300

30

0.3

Setting H

\(\frac {1}{3}\)

300

30

0.5

Setting I

\(\frac {1}{3}\)

300

30

\(\frac {1}{30}, \frac {2}{30}, \dots, 1\)

Setting J

\(\frac {1}{3}\)

300

30

\(\frac {1}{30}, \frac {2}{30}, \dots, 1\)

Setting K

\(\frac {1}{3}\)

300

30

0.5

  1. For every setting B=100 training data sets are generated. Settings G to K are specialized settings, which focus on changes in the data generation process. For details see “Settings with altered simulation design” section