Table 2

Features sets, naïve Bayesian classifier

Set Size

Features Selected


Trial 1

Trial 2


1

96

96

2

87

87

3

90

90

4

11

32

5

54

54

6

40

11

7

15

6

8

46

35

9

21

60

10

48

53


Two typical trials of a naïve Bayesian classifier wrapper. After the first three features, selection is very unstable. Similar results occurred for all the traditional classifiers.

Kuschner et al. BMC Bioinformatics 2010 11:177   doi:10.1186/1471-2105-11-177

Open Data