Table 2.
Predicted performances of classifiers trained with 1,742 examples, presented as the mean and standard deviation (Mean ± SD)
| ALG | FS | Acc | Se | Sp | Fm | Mcc |
|---|---|---|---|---|---|---|
| SVM |
FS4 |
E 85.6 ± 1.2 a |
D 83.0 ± 1.9 a |
D 88.4 ± 1.5 a |
E 85.2 ± 1.3 a |
E 71.4 ± 2.3 a |
| |
FS5 |
D 87.4 ± 0.9 a |
C 84.3 ± 1.5 a |
C 90.5 ± 1.4 a |
D 86.9 ± 0.9 a |
D 74.9 ± 1.7 a |
| |
FS6 |
C 89.8 ± 1.1 a |
B 87.5 ± 1.5 a |
C 93.0 ± 1.7 a |
C 89.5 ± 1.1 a |
C 79.8 ± 2.2 a |
| |
FS3 |
B 90.6 ± 0.8 a |
B 88.0 ± 1.3 a |
B 93.3 ± 1.3 a |
B 90.4 ± 0.9 a |
B 81.4 ± 1.7 a |
| |
FS1 |
A 92.2
±
0.9 a |
A 89.7
±
1.8 a |
A 94.7
±
0.8 a |
A 92.0
±
1.0 a |
A 84.6
±
1.8 a |
| |
FS2 |
A 92.4
±
0.9 a |
A 90.1
±
1.6 a |
A 94.7
±
0.6 a |
A 92.2
±
1.0 a |
A 84.9
±
1.8 a |
| |
FS7 |
A 92.3
±
1.0 a |
A 89.9
±
1.1 a |
A 94.7
±
0.9 a |
A 92.1
±
0.9 a |
A 84.7
±
1.6 a |
| |
SELECT |
A 92.3
±
0.9 a |
A 90.0
±
1.3 a |
A 94.6
±
1.0 a |
A 92.1
±
0.9 a |
A 84.6
±
1.7 a |
| RF |
FS4 |
E 84.8 ± 1.1 b |
D 81.2 ± 1.8 b |
C 88.3 ± 1.3 a |
E 84.2 ± 1.2 b |
E 69.8 ± 2.1 b |
| |
FS5 |
D 85.7 ± 0.7 b |
D 81.2 ± 0.8 b |
B 90.3 ± 1.4 a |
D 85.1 ± 0.6 b |
D 71.8 ± 1.5 b |
| |
FS6 |
C 88.7 ± 1.4 b |
C 86.6 ± 1.5 b |
A 89.8 ± 1.6 b |
C 88.5 ± 1.4 b |
C 77.4 ± 2.8 b |
| |
FS3 |
C 90.0 ± 1.0 b |
C 86.9 ± 1.4 b |
A 93.0 ± 1.1 a |
C 89.6 ± 1.0 b |
C 80.1 ± 1.9 b |
| |
FS1 |
A 91.5 ± 1.0 b |
A 89.1 ± 1.1 a |
A 93.9 ± 1.2 a |
A 91.3 ± 1.0 b |
A 83.1 ± 1.9 b |
| |
FS2 |
A 90.9 ± 1.0 b |
B 88.1 ± 1.2 b |
A 93.8 ± 1.3 b |
A 90.7 ± 1.1 b |
A 82.0 ± 2.1 b |
| |
FS7 |
A 91.1 ± 0.8 b |
B 88.5 ± 1.3 b |
A 93.7 ± 1.3 b |
A 90.9 ± 1.0 b |
A 82.3 ± 2.0 b |
| |
SELECT |
B 90.5 ± 0.9 b |
C 87.4 ± 1.0 b |
A 93.6 ± 1.4 b |
B 90.2 ± 0.9 b |
B 81.2 ± 1.9 b |
| G 2DE | FS3 | 90.2 ± 0.9 | 87.4 ± 1.5 | 93.1 ± 0.9 | 89.9 ± 0.9 | 80.6 ± 1.8 |
Predicted accuracies (Acc), sensitivities (Se), specificities (Sp), F-measures (Fm) and Mathew Correlation Coefficients (Mcc) of classifiers trained with 1,742 examples, presented as the mean and standard deviation (mean ± sd). Capital letters in columns indicate the performance cluster of each feature set, within algorithm (ALG). Lower case letters in columns indicate the cluster of each algorithms, within feature sets. Bold numbers represents the highest performances, which were not significantly different according to the clustering criteria in [42].