CODEWITHSUNDEEP
machine-learningclassificationwekadata-miningroc
manu
manu

Reputation: 107

WEKA ROC CURVE - colour interpretations

When we generate ROC curves from Weka, curve is displayed in a color scale. I understand what is a ROC curve. But what is the meaning of the color scale in this ROC curve?

enter image description here

Upvotes: 3

Views: 1622

Answers (1)

lejlot
lejlot

Reputation: 66815

Color depends on what have you chosen. In your case, you chosen Colour: Threshold, so the color represents threshold value set to get this pair of true FPR/TPR point.

In other words - ROC is a parametric curve of the threshold, so point (tpr, fpr) belongs to ROC if and only if there exists threshold value t for which experiment results in True positive rate=tpr and False positive rate=fpr. However, it is just an existance, so ROC as such does not show what is the exact value of t used. Coloring the curve gives you this one missing information.

Upvotes: 4

Related Questions