Conformal Regression on Motorcycle accident n = 133 (50/50 splits)
93.1CoverageCQR-QRF
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CQR-QRFBase regressor=QRF fitted on the training half, conformalized on the calibration half, Nominal level=1 - ε = 0.90, Number of splits=2002026.03 | 93.1 | 87.9 | |
| Gaussian split conformalNominal level=1 - ε = 0.90, Number of splits=2002026.03 | 92.5 | 172.4 | |
| CQR (cubic)Nominal level=1 - ε = 0.90, Number of splits=2002026.03 | 92.5 | 134.1 | |
| CRPS-Optimal BinningFitting strategy=uses all n observations for fitting, Nominal level=1 - ε = 0.90, Number of splits=2002026.03 | 91 | 78.9 | |
| CRPS-Optimal BinningFitting strategy=uses only the training half, Nominal level=1 - ε = 0.90, Number of splits=2002026.03 | 87 | 100.5 |