Table 4.
Confusion matrices and performance metrics for ESI+ and ESI− downsampled model predictions compared to external validation data
| Amenable (prediction) |
Unamenable (prediction) |
|
|---|---|---|
| ESI− downsampled model | ||
| Detected (experiment) | 323 | 502 |
| Not-detected (experiment) | 68 | 874 |
| Sensitivity | 0.83 | |
| Specificity | 0.64 | |
| Balanced accuracy | 0.73 | |
| ESI+ downsampled model | ||
| Detected (experiment) | 423 | 402 |
| Not-detected (experiment) | 103 | 839 |
| Sensitivity | 0.80 | |
| Specificity | 0.68 | |
| Balanced accuracy | 0.74 | |
| Combined models | ||
| Detected (experiment) | 505 | 320 |
| Not-detected (experiment) | 129 | 813 |
| Sensitivity | 0.80 | |
| Specificity | 0.72 | |
| Balanced accuracy | 0.76 |