Table 3.
Predictive performance of the machine learning methods per-class statistics based on under-sampling.
| Performance measures | Methods | ||||||
|---|---|---|---|---|---|---|---|
| Class | SVM-R | SVM-L | SVM-P | ANN | kNN | Bagging trees | |
| Accuracy | BRCA | 98.6 | 97.3 | 99.2 | 87.7 | 96.0 | 99.5 |
| COAD | 95.8 | 98.6 | 98.6 | 90.2 | 94.7 | 98.5 | |
| LUAD | 97.7 | 99.6 | 98.0 | 82.8 | 90.6 | 98.7 | |
| OV | 90.7 | 88.5 | 98.9 | 93.4 | 98.5 | 100 | |
| THCA | 97.8 | 100 | 100 | 82.5 | 99.1 | 99.6 | |
| Sensitivity | BRCA | 99.4 | 100 | 99.7 | 84.8 | 92.7 | 99.7 |
| COAD | 91.7 | 97.2 | 97.2 | 86.1 | 94.4 | 97.2 | |
| LUAD | 98.8 | 100 | 96.5 | 68.6 | 81.4 | 97.7 | |
| OV | 81.6 | 77.0 | 97.7 | 92.0 | 98.9 | 100 | |
| THCA | 95.5 | 100 | 100 | 67.6 | 98.2 | 99.1 | |
| Specificity | BRCA | 97.8 | 94.7 | 98.8 | 90.6 | 99.4 | 99.4 |
| COAD | 100 | 100 | 100 | 94.3 | 94.9 | 99.8 | |
| LUAD | 96.6 | 99.3 | 99.5 | 97.0 | 99.8 | 99.6 | |
| OV | 99.8 | 100 | 100 | 94.8 | 98.0 | 100 | |
| THCA | 100 | 100 | 100 | 97.4 | 100 | 100 | |
| F1-score | BRCA | 98.6 | 97.5 | 99.2 | 87.5 | 95.9 | 99.5 |
| COAD | 95.7 | 98.6 | 98.6 | 60.8 | 67.3 | 97.2 | |
| LUAD | 89.5 | 97.7 | 96.5 | 72.8 | 89.2 | 97.7 | |
| OV | 89.3 | 87.0 | 98.8 | 81.6 | 93.5 | 100 | |
| THCA | 97.7 | 100 | 100 | 75.0 | 99.1 | 99.6 | |
| Precision | BRCA | 97.9 | 95.1 | 98.8 | 90.3 | 99.4 | 99.4 |
| COAD | 100 | 100 | 100 | 47.0 | 52.3 | 97.2 | |
| LUAD | 81.7 | 95.6 | 96.5 | 77.6 | 98.6 | 97.7 | |
| OV | 98.6 | 100 | 100 | 73.4 | 88.7 | 100 | |
| THCA | 100 | 100 | 100 | 84.3 | 100 | 100 | |
SVM-R support vector machine with radial-basis function (RBF) kernel, SVM-L support vector machine with linear kernel, SVM-P support vector machine with polynomial kernel, ANN Artificial Neural Networks, kNN K-nearest neighbors, Bagging trees, ACC Accuracy, CI confidence interval, Kappa kappa statistics AUC area under the curve.