Table 2.
Stratified 10-fold cross-validation results with ECFP4 over ten iterations.
| Iteration | ACC | Best Loss |
|---|---|---|
| 1 | 0.873 ± 0.0957 | 0.2271 |
| 2 | 0.864 ± 0.1074 | 0.1713 |
| 3 | 0.945 ± 0.0834 | 0.1264 |
| 4 | 0.918 ± 0.0725 | 0.1066 |
| 5 | 0.939 ± 0.0778 | 0.1276 |
| 6 | 0.940 ± 0.0859 | 0.0837 |
| 7 | 0.835 ± 0.1557 | 0.1858 |
| 8 | 0.927 ± 0.0788 | 0.1102 |
| 9 | 0.889 ± 0.0943 | 0.1258 |
| 10 | 0.814 ± 0.1928 | 0.2245 |
ACC: accuracy; DILI+: drug-induced liver injury (DILI)-positive; DILI−: DILI-negative. Bold represents the best model selected due to the lowest loss values.