Table 1.
Performance of RF classifiers with different combinations of feature types under the entire tenfold cross-validation#.
| Model | Accuracy | Precision | Recall | F1-measure | MCCa |
|---|---|---|---|---|---|
| Addition | 89.26% | 92.53% | 86.85% | 89.60% | 78.69% |
| Subtraction | 88.84% | 92.31% | 86.32% | 89.22% | 77.88% |
| Hadamard | 73.63% | 74.71% | 73.13% | 73.91% | 47.27% |
| Addition + subtraction | 89.33% | 92.70% | 87.20% | 89.86% | 79.25% |
| Addition + Hadamard | 88.51% | 92.14% | 86.64% | 89.30% | 78.09% |
| Subtraction + Hadamard | 88.81% | 92.01% | 86.48% | 89.16% | 77.78% |
| Addition + subtraction + Hadamard | 89.47% | 92.34% | 87.33% | 89.76% | 79.08% |
#Numbers in italics indicate the highest values in the corresponding column. aThe Mathews correlation coefficient.