Table 2:
Performance of Different Classification Algorithms with all Variables.
| Classifier | Ten-fold cross validation performance (%) |
||||
|---|---|---|---|---|---|
| P | R | F | AUC | P-value* | |
| BASELINE | 61.68 | 61.58 | 61.54 | 50.64 | 1.06E-9 |
| Logistic Regression | 70.82 | 92.02 | 80.04 | 72.78 | 2.85E-1 |
| SVM + Linear Kernel | 70.22 | 92.02 | 79.65 | 69.91 | 2.83E-1 |
| SVM + RBF Kernel | 70.35 | 93.12 | 80.15 | 69.46 | N/A |
| Random Forest | 72.52 | 79.31 | 75.76 | 72.13 | 5.56E-6 |
| Classifier | Test set performance (%) |
||||
|---|---|---|---|---|---|
| P | R | F | AUC | ||
| BASELINE | 60.70 | 59.51 | 60.10 | 50.65 | |
| Logistic Regression | 71.52 | 92.68 | 80.74 | 75.47 | |
| SVM + Linear kernel | 70.52 | 92.20 | 79.92 | 68.07 | |
| SVM + RBF kernel | 69.46 | 93.17 | 79.58 | 70.58 | |
| Random Forest | 72.25 | 80.00 | 75.93 | 72.96 | |
Bold numbers indicate the best results.
*The P-value was calculated by comparing the F-measure between the best algorithm (SVM + RBF kernel) and the other algorithms using the paired t-test in 10-fold cross-validation.
N/A indicates that the performances between the two algorithms are identical and no P-value is returned.