Table 3.
The classification results before and after feature selection. FS represents the method after feature selection. MAUC represents multi-class area under the curve. The statistical tests using the Mann Whitney U Test were performed between the overall accuracies using methods without feature selection and those with feature selection.
| Classifier | Overall accuracy (%) | Balanced accuracy (%) | MAUC (%) |
|---|---|---|---|
| SVM | 51.9 ± 1.1 | 59.5 ± 1.4 | 81.8 ± 0.7 |
| RF | 73.8 ± 0.4 | 49.6 ± 0.7 | 86.2 ± 0.5 |
| SMOTE | 71.0 ± 0.6 | 60.1 ± 1.0 | 85.0 ± 1.1 |
| mckNN | 69.8 ± 0.7 | 57.7 ± 1.1 | – |
| RUSBoost | 72.5 ± 0.8 | 66.2 ± 0.9 | 88.1 ± 0.3 |
| SVM_FS* | 65.4 ± 1.4 | 55.6 ± 1.9 | 84.7 ± 0.9 |
| RF_FS* | 75.2 ± 0.6 | 52.2 ± 0.9 | 87.9 ± 0.5 |
| SMOTE_FS* | 73.1 ± 1.1 | 63.8 ± 1.6 | 86.7 ± 1.0 |
| mckNN_FS* | 73.4 ± 1.1 | 60.0 ± 1.7 | – |
| RUSBoost_FS* | 75.2 ± 0.8 | 69.3 ± 1.0 | 89.3 ± 0.5 |
*
Means that the results with feature selection are significantly different from those without feature selection with p-value <0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)