MACHINE LEARNING / EVALUATION
Model Evaluation Lab
Two sets of supplied synthetic scores. One shared label set. Thresholds, ranking, and prediction ensembles in comparison.
Loading evaluation data…
Positive when score ≥ threshold. Mean ensemble score = (A + B) / 2.
Accuracy—Threshold dependent
F1—Positive-class F1
ROC AUC—Ranking across thresholds
Average precision—PR summary, not trapezoidal AUC
Mean individual accuracy—(Accuracy A + Accuracy B) / 2
Accuracy of mean predictions—Evaluate (score A + score B) / 2
Different operations can produce different results; ensemble gains are not guaranteed.
Same-threshold comparison
| Model | Accuracy | Precision | Recall | F1 |
|---|
Ranking curves
The larger dot marks the displayed model at the selected threshold. Full curves and AUC / AP use every unique score boundary.
The PR curve ends at recall 0, precision 1 by convention; the threshold marker uses precision 0 when no positives are predicted.
Confusion matrix
Rows = true labels. Columns = predictions.
| True / Predicted | Negative (0) | Positive (1) |
|---|---|---|
| Negative (0) | TN— | FP— |
| Positive (1) | FN— | TP— |
Sample predictions
| ID | True | A | B | Mean | Prediction | Result |
|---|