How to interpret thresholds of ROC AUC Curves for Decision Trees?

Question

When using a discrete classifier like decision tree, we get a single point (FPR, TPR) by through the confusion matrix, now when I try to plot ROC AUC curve, I get thresholds :

roc_curve(y_test,mod.predict(X_test))

Output :

(array([  0.00000000e+00,   5.92624518e-04,   1.00000000e+00]),
 array([ 0.        ,  0.11766772,  1.        ]),
 array([ 2.,  1.,  0.]))

threshold = [2.,1.,0.,]

I am unable to interpret these thresholds, how do I interpret them to find TPR and FPR?

Please update your question to include which package(s) you are using. IE where is roc_curve from? — FiddleStix

FEldin FEldin · Accepted Answer · 2020-08-14T18:37:54

Look here sklearn.metrics.roc_curve.html

The first array of your roc_curve return is fpr, the second is tpr and the thirs is threshold
when you plot fpr (X) on tpr (Y) for each threshold yo get the ROC curve

How to interpret thresholds of ROC AUC Curves for Decision Trees?

1 Answers