Table 1. The effect of lpICS reweighting on the breast cancer dataset.
Complexity is reported as the number of selected variables and the total number of variable intervals. Training and test results are given as the Area Under the Curve (AUC), the accuracy (acc) and the slope and bias of the calibration curve.
| Without reweighting | |||
| 23 vars | 68 intervals | ||
| Training results | |||
| AUC | 1 | Acc | 1 |
| Slope | 1 | Bias | 0 |
| Test results | |||
| AUC | 0.989 | Acc | 0.890 |
| Slope | 0.762 | Bias | 0.019 |
| With reweighting | |||
| 2 vars | 5 intervals | ||
| Training results | |||
| AUC | 0.958 | Acc | 0.958 |
| Slope | 1 | Bias | 0 |
| Test results | |||
| AUC | 0.933 | Acc | 0.916 |
| Slope | 0.893 | Bias | 0.030 |