Recursive Feature Selection with Significant Variables of Support Vectors
Table 9
The CV results on the lung cancer dataset.
Num. of genes
RSVM
SVM-RFE
SVM-
CV2
Mean SV
CV2
Mean SV
CV2
Mean SV
51.34
51.34
51.34
6060
46.78
51.35
50.85
5151
45.07
51.40
50.20
4378
44.24
51.42
49.24
3721
43.99
51.35
48.22
3163
43.53
51.34
47.60
2689
43.38
51.36
47.38
2286
43.12
51.86
46.71
1943
43.17
52.79
46.05
1652
43.26
52.51
45.10
1404
43.12
51.84
43.78
1193
42.87
52.33
42.60
1014
42.86
51.15
41.84
862
42.24
51.64
41.50
733
41.47
52.21
41.03
623
41.55
51.72
40.07
530
40.85
51.28
39.71
450
40.40
50.59
39.20
382
39.08
50.70
38.34
325
38.62
50.77
37.43
276
38.52
49.21
36.62
235
38.71
48.27
36.06
200
37.28
47.67
35.59
170
36.33
45.84
35.13
144
36.19
44.64
33.88
122
35.43
42.81
33.48
104
34.30
40.87
32.59
88
32.87
39.02
31.99
75
31.28
36.56
31.92
64
29.84
34.08
31.80
54
27.41
30.07
31.97
46
25.94
27.50
31.90
39
24.42
25.10
32.30
33
22.53
23.09
32.66
28
20.42
19.87
32.97
24
18.86
19.31
33.16
20
16.94
16.59
33.51
17
15.83
15.71
33.58
14
16.42
16.15
34.08
12
17.72
17.03
34.55
10
19.76
19.95
35.15
8
22.45
22.60
36.01
7
23.64
24.16
36.44
6
25.35
25.97
36.87
5
27.86
28.73
37.50
Num. of genes: the number of selected genes in the recursive steps. CV2: total cross-validation error rate (CV2 error rate). Mean SV: mean number of support vectors used in the cross-validation at each level.