Table 5.
Estimated performance of the different combination feature sets/classifiers and pairwise differences. Key to symbols: “ACC” = accuracy, “SP” = specificity, “SN” = sensitivity. Values are in %, differences in percentage points. Boldface figures indicate significant differences. For a comparison: accuracy of a random classifier (blind to prior class probabilities) = 50%; with prior class probabilities = 54.4%.
| Model | Classifier | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ClT | KNN | NBGaussian | |||||||||
| ACC | SN | SP | ACC | SN | SP | ACC | SN | SP | |||
| CTb | 58.8 | 67.6 | 42.2 | 59.5 | 65.9 | 47.5 | 69.5 | 83.5 | 43.3 | ||
| CTe | 62.2 | 69.6 | 48.2 | 70.7 | 78.7 | 55.8 | 74.3 | 86.9 | 50.8 | ||
| PETb | 73.4 | 78.7 | 63.5 | 72.5 | 76.9 | 64.4 | 72.2 | 69.9 | 76.6 | ||
| PETe | 75.7 | 82.1 | 63.6 | 77.1 | 81.4 | 69.1 | 82.4 | 87.6 | 72.6 | ||
| CTb+PETb | 71.2 | 77.1 | 60.1 | 74.4 | 78.9 | 66.1 | 70.6 | 71.3 | 69.1 | ||
| CTe+PETe | 72.3 | 78.9 | 59.8 | 73.7 | 75.7 | 70.1 | 80.4 | 88.5 | 65.2 | ||
| CTe-CTb | + 3.4 | +2.0 | + 6.0 | + 11.2 | + 12.8 | + 8.3 | + 4.8 | + 3.4 | + 7.5 | ||
| PETe-PETb | + 2.2 | + 3.4 | +0.1 | + 4.6 | + 4.5 | + 4.7 | + 10.2 | + 17.7 | −3.9 | ||