Predictive process monitoring on Sepsis (temporal split)
91.3AccuracyTwo-Stage-T
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Two-Stage-TEvaluation protocol=Temporal split, Seeds=5, Architecture=Transformer-based with weighted axiom loss and rule pruning2026.03 | 91.3 | 75.63 | |
| TRFEvaluation protocol=Temporal split, Seeds=5, Architecture=Transformer (2 attention heads, hidden 128)2026.03 | 90.11 | 73.8 | |
| Two-Stage-LEvaluation protocol=Temporal split, Seeds=5, Architecture=LSTM-based with weighted axiom loss and rule pruning2026.03 | 89.39 | 75.82 | |
| LSTMEvaluation protocol=Temporal split, Seeds=5, Architecture=LSTM (2 layers, hidden 128)2026.03 | 88.69 | 70.59 |