Table 3.
Comparison of models for the CFSAN dataset
|
External validation using the FDA/CFSAN dataset (1399 structures) | |||||
|---|---|---|---|---|---|
| BAC | SEN | SPEC | PPV | NPV | |
| SOHN |
0.75 |
0.64 |
0.86 |
0.75 |
0.79 |
| Random Forest |
0.77 |
0.74 |
0.80 |
0.70 |
0.83 |
| SVM |
0.78 |
0.70 |
0.85 |
0.74 |
0.82 |
| kNN |
0.73 |
0.62 |
0.83 |
0.70 |
0.78 |
| Decision Tree | 0.74 | 0.62 | 0.87 | 0.74 | 0.78 |
Results for the confidential dataset. BAC = balanced accuracy, SEN = sensitivity, SPEC = specificity, PPV = positive predictivity, NPV = negative predictivity.