Table 1. Statistics of the prediction performances.
Model | SE (RF/SVM) | SP (RF/SVM) | CO (RF/SVM) | AUC (RF/SVM) | ||||
Training | Test | Training | Test | Training | Test | Training | Test | |
Model I | 79.58%/76.31% | 80.99%/77.50% | 84.11%/85.85% | 84.66%/85.91% | 81.84%/81.08% | 82.83%/81.71% | 89.50/88.97 | 90.55/89.91 |
Model II | 81.18%/77.66% | 73.66%/74.11% | 83.89%/86.11% | 83.52%/85.49% | 82.50%/81.78% | 79.09%/80.38% | 90.31/89.79 | 86.50/87.19 |
Model III | 81.33%/77.56% | 52.20%/55.17% | 83.22%/85.35% | 92.15%/93.62% | 82.25%/81.34% | 74.53%/76.66% | 89.81/89.04 | 82.77/84.90 |
Model IV | 71.90%/66.41% | 32.52%/36.48% | 90.25%/88.07% | 91.21%/92.42% | 82.46%/78.88% | 66.09%/68.48% | 88.99/85.62 | 72.64/75.47 |
Average | 78.50%/74.49% | 59.84%/60.82% | 85.37%/86.35% | 87.89%/89.36% | 82.26%/80.77% | 75.64%/76.81% | 89.65/88.36 | 83.13/83.93 |
The AUC (ROC score) is the area under the ROC curve, normalized to 100 for a perfect inference and 50 for a random inference.