Table 13.
Best results for THRESHOLD 0.65: We counted the number of times the sensitivity threshold of 0.65 was exceeded. We considered variants v1 to v6. The best method was found to be svm_linear and com3 and variants v4 and v2. Slightly better were v4 and svm_linear. It is worth noting that com3 also includes svm_linear. In yellow, we highlighted the best methods and variants.
| Method | sum(:) | sum(:) | ||||||
|---|---|---|---|---|---|---|---|---|
| 1nn | 0 | 1 | 0 | 1 | 1 | 0 | 2 | 3 |
| nb | 0 | 0 | 0 | 1 | 0 | 0 | 1 | 1 |
| svm_linear | 2 | 5 | 3 | 4 | 4 | 3 | 14 | 21 |
| svm_radial | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| svm_polynomial | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| svm_sigmoid | 0 | 0 | 0 | 1 | 0 | 0 | 1 | 1 |
| lg | 1 | 0 | 2 | 2 | 0 | 0 | 5 | 5 |
| rf | 1 | 1 | 0 | 1 | 1 | 0 | 3 | 4 |
| com3 | 0 | 5 | 3 | 5 | − | − | 13 | − |
| sum | 4 | 12 | 8 | 15 | 6 | 3 |