Research Article
Mathematical Programming Approaches to Classification Problems
Table 11
The SVM-based approach results.
| | | D1 | D2 | D3 | D4 |
| | CPU-time average | 28 seconds | 78 seconds | 54 seconds | 200 seconds | | Apparent hit rates | 91,3 () | 75,8 () | 92 () | 97,2 () | | Linear SVM | Training apparent hit rates | 95,8 () | 73,3 () | 87,1 () | 94,63 () | | Holdout sample hit rates | 86,36 () | 64,7 () | 80 () | 97,8 () |
| | CPU-time average | 29 seconds | 53 seconds | 51 seconds | 330 seconds | | | radial: | radial: | radial: | radial: | | Kernel Function for the | C=100 | c=1000 | c=1000 | c=100 | | complete dataset | e= 0,01 | e=1e-006 | e=1e-005 | e=0,2 | | | gamma = 10 | gamma = 10 | gamma = 2 | Gamma = 5 | | Apparent hit rates | 100 (0) | 100 (0) | 100 (0) | 100 (0) | | Nonlinear SVM | | Anova: | Anova: | Polynomial: | Polynomial: | | | c=100000 | c=10 | c=100 | c=100 | | Kernel Function for | e= 0,01 | e=0,1 | e=0,2 | e=0 | | training dataset | gamma = 0,1 | gamma = 0,9 | degree = 2 | degree = 3 | | | degree = 2 | degree = 4 | | | | Training apparent hit rates | 100 (0) | 100 (0) | 100 (0) | 100 (0) | | Holdout sample hit rates | 86,36 () | 70,58 () | 80 () | 96,17 () |
|
|
