Chosen model
gradient_boosting
CV ROC-AUC (mean ± std)
0.9457 ± 0.0488
Test ROC-AUC
1.000
Test PR-AUC
1.000
Train size
3815
Evaluation
| split | ROC-AUC | PR-AUC | precision | recall | F1 | pos / neg |
|---|---|---|---|---|---|---|
| validation | 0.917 | 0.587 | 80.0% | 44.4% | 0.571 | 9 / 1263 |
| test | 1.000 | 1.000 | 100.0% | 60.0% | 0.750 | 5 / 565 |
Confusion matrix (test, threshold 0.5)
| predicted non-exoplanet | predicted exoplanet | |
|---|---|---|
| actual non-exoplanet | 565 | 0 |
| actual exoplanet | 2 | 3 |
Curves
Training curve
Validation ROC-AUC measured after every 10th boosting iteration of the chosen model; the last point is the fitted model.
What the model looks at
Importance from gradient_boosting.