Table 5:
Comparison of the classifiers using a paired two-tailed t-test.
| B-LR with Gaussian priors versus each of the following: | Accuracy | BACC | AUC |
|---|---|---|---|
| B-LRGP2 | 0.18 (1.35) | 0.73(0.31) | 0.14(2.37) |
| B-LRLP | 0.01(4.39) | 0.10(3.06) | 0.01(7.42) |
| J48 | 0.01(12.63) | 0.00(18.13) | 0.02(13.11) |
| NB | 0.02(3.78) | 0.01(6.13) | 0.98(−0.08) |
| LRsimple | 0.06(4.22) | 0.02(7.8) | 0.86(−0.48) |
| LRridge | 0.20(1.99) | 0.06(3.18) | 0.04(−4.09) |
| CART | 0.00(10.93) | 0.00(23.5) | 0.01(19.48) |
| SVM | 0.06(−0.65) | 0.57(−0.19) | 0.02(−1.29) |
| RF | 0.01(5.71) | 0.04(10.00) | 0.55(−1.3) |
The numbers shown are p-values; the values below 0.05 are shown in bold. The value in a parenthesis is the mean performance of B-LR minus the mean performance of the listed classifier, expressed as a percentage. The values underlined are those in which the p-value is less than 0.05 and B-LR performed better.