Train mode
ensemble-kfold
Latest PhaseB Validation
This page intentionally uses the most recent PhaseB validation run only: ffnn_ens_2.4_v100_28234651 (last modified 2026-04-30). Multi-dataset PhaseC validation is intentionally excluded until stability is established.
Split type
id-family
Runs / Folds
5 / 5
PR baseline mean
6.968%
Split seed(s): 101 | Train seed(s): 11 | Best model metric: loss
Run-Level Summary (n=5 folds)
| Metric | Mean | Std | Min | Max |
|---|---|---|---|---|
| PR-AUC | 0.246263 | 0.024043 | 0.212074 | 0.272028 |
| F1 | 0.317403 | 0.044387 | 0.275047 | 0.371690 |
| MCC | 0.267715 | 0.036428 | 0.235543 | 0.327421 |
| ROC AUC | 0.787045 | 0.037246 | 0.744557 | 0.831081 |
| AUC10 | 0.275223 | 0.079704 | 0.146739 | 0.365858 |
| Balanced accuracy | 0.672451 | 0.046903 | 0.629480 | 0.744991 |
Fold-Level Best Checkpoints
| Fold | Best epoch | PR-AUC | F1 | MCC | AUC | AUC10 | Threshold |
|---|---|---|---|---|---|---|---|
| 1 | 8 | 0.257702 | 0.358406 | 0.327421 | 0.811175 | 0.277892 | 0.473966 |
| 2 | 1 | 0.272028 | 0.371690 | 0.267417 | 0.744557 | 0.146739 | 0.316929 |
| 3 | 10 | 0.257872 | 0.296305 | 0.267082 | 0.831081 | 0.365858 | 0.758105 |
| 4 | 23 | 0.212074 | 0.275047 | 0.235543 | 0.753081 | 0.298390 | 0.571140 |
| 5 | 11 | 0.231639 | 0.285568 | 0.241113 | 0.795333 | 0.287239 | 0.712729 |
