Survival Prediction and Feature Selection in Patients with Breast Cancer Using Support Vector Regression
Table 2
Performance measures of Cox and SVR models using different kernels and datasets when all features are included in the model. Statistical significant differences between SVR based model 1 (indicated in >italic) and the other models are indicated based on the Wilcoxon rank sum test.
Dataset
Model
Type of kernel
c-index
Log rank test statistic
Hazard ratio
BC
Cox model
—
0.61 ± 0.03
1.38 ± 1.10
1.22 ± 0.11
SVR based model 1
Linear
0.61 ± 0.03
1.85 ± 1.48
1.33 ± 0.14
SVR-MRL based model 1
Linear
0.62 ± 0.03
1.95 ± 1.44
1.33 ± 0.15
SVR based model 2
Linear
0.60 ± 0.03
1.76 ± 1.33
1.31 ± 0.14
SVR based model 1
RBF
0.56 ± 0.03
0.52 ± 0.46
1.13 ± 0.07
SVR based model 2
RBF
0.54 ± 0.05
0.64 ± 0.64
1.14 ± 0.11
SVR based model 1
Polynomial
0.59 ± 0.04
1.56 ± 1.56
1.23 ± 0.17
SVR based model 2
Polynomial
0.57 ± 0.08
0.88 ± 0.88
1.19 ± 0.23
SVR based model 1
Clinical
0.60 ± 0.02
1.47 ± 1.11
1.26 ± 0.14
SVR based model 2
Clinical
0.60 ± 0.04
1.14 ± 0.97
1.27 ± 0.14
CD
Cox model
—
0.64 ± 0.01
20.81 ± 5.33
1.53 ± 0.07
SVR based model 1
Linear
0.64 ± 0.01
21.59 ± 5.53
1.49 ± 0.07
SVR based model 2
Linear
0.64 ± 0.01
22.17 ± 6.41
1.51 ± 0.07
SVR-MRL based model 1
Linear
0.66 ± 0.01
22.60 ± 5.51
1.50 ± 0.09
SVR based model 1
RBF
0.61 ± 0.01
12.13 ± 3.49
1.46 ± 0.08
SVR based model 2
RBF
0.61 ± 0.01
11.98 ± 3.98
1.43 ± 0.07
SVR based model 1
Polynomial
0.63 ± 0.01
18.94 ± 7.05
1.56 ± 0.11
SVR based model 2
Polynomial
0.63 ± 0.02
14.72 ± 9.65
1.54 ± 0.17
SVR based model 1
Clinical
0.65 ± 0.01
22.63 ± 4.80
1.63 ± 0.07
SVR based model 2
Clinical
0.65 ± 0.01
22.28 ± 5.12
1.65 ± 0.07
PT
Cox model
—
0.70 ± 0.05
2.13 ± 1.50
1.41 ± 0.24
SVR based model 1
Linear
0.74 ± 0.06
3.18 ± 2.24
2.08 ± 0.49
SVR based model 2
Linear
0.74 ± 0.06
3.14 ± 1.92
2.14 ± 0.48
SVR-MRL based model 1
Linear
0.76 ± 0.06
3.91 ± 2.40
2.12 ± 0.49
SVR based model 1
RBF
0.68 ± 0.06
0.81 ± 0.75
1.45 ± 0.30
SVR based model 2
RBF
0.67 ± 0.05
0.84 ± 0.74
1.45 ± 0.28
SVR based model 1
Polynomial
0.74 ± 0.05
3.69 ± 2.03
2.15 ± 0.52
SVR based model 2
Polynomial
0.76 ± 0.06
3.76 ± 2.32
2.35 ± 0.49
SVR based model 1
Clinical
0.71 ± 0.07
1.85 ± 1.67
1.83 ± 0.46
SVR based model 2
Clinical
0.70 ± 0.07
1.60 ± 1.42
1.83 ± 0.48
PD
Cox model
—
0.82 ± 0.02
23.11 ± 5.24
2.67 ± 0.31
SVR based model 1
Linear
0.84 ± 0.01
26.31 ± 5.19
3.12 ± 0.55
SVR based model 2
Linear
0.83 ± 0.01
27.40 ± 5.59
3.07 ± 0.54
SVR-MRL based model 1
Linear
0.84 ± 0.01
26.09 ± 5.82
3.14 ± 0.52
SVR based model 1
RBF
0.84 ± 0.02
27.93 ± 4.46
3.02 ± 0.55
SVR based model 2
RBF
0.84 ± 0.02
28.51 ± 4.61
3.01 ± 0.59
SVR based model 1
Polynomial
0.84 ± 0.02
26.58 ± 4.52
3.02 ± 0.52
SVR based model 2
Polynomial
0.84 ± 0.02
26.61 ± 4.85
3.12 ± 0.49
SVR based model 1
Clinical
0.83 ± 0.01
23.92 ± 4.80
3.21 ± 0.56
SVR based model 2
Clinical
0.83 ± 0.01
25.11 ± 5.23
3.14 ± 0.54
value < 0.05, value < 0.01, value < 0.001 (Wilcoxon rank sum test).