ROC from Probabilities¶
Computes Receiver operating characteristics (ROC) based on calculated Y-probabilities and from true Y.
Documentation¶
The ROC (receiver operating characteristic) curve is a graphical plot which illustrates the performance of a binary classifier system as its discrimination threshold is varied. It is created by plotting the fraction of true positives out of the positives (TPR = true positive rate) vs. the fraction of false positives out of the negatives (FPR = false positive rate), at various threshold settings.
The threshold will typically vary from 1 (or infinity), where only classifications with 100% confidence will be considered as correct, to 0 where all classifications will be considered correct. Thus the curve will always go from (TPR=0.0, FPR=0.0) to (TPR=1.0, FPR=1.0). A good classifier will have a high TPR for all thresholds, while a bad (random) classifier will yield a linear roc curve from (0, 0) to (1, 1).
Definition¶
Input ports¶
- Y-prob
- Y-true
Output ports¶
- roc
Configuration¶
- Drop suboptimal thresholds (drop_intermediate)
Whether to drop some suboptimal thresholds which would not appear on a plotted ROC curve. This is useful in order to create lighter ROC curves.
Added in version 0.17: parameter drop_intermediate.
- header as label (header as label)
Use header of Y-prob as the target label
- Positive class label (pos_label)
The label of the positive class. When
pos_label=None, if y_true is in {-1, 1} or {0, 1},pos_labelis set to 1, otherwise an error will be raised.
Examples¶
The node can be found in:
Implementation¶
- class node_metrics.ROCFromProb[source]
