Hello!
I have a suggestion to improve the ‘Classification Threshold Analysis’ component to enhance its suitability for calculating PR curve and RP curve AUC
I found " Classification Threshold Analysis" component very useful. Thanks and respect to the author ( mlauber71).
And as mlauber71 suggests in
Precision Recall Curve and AUC,
I am trying to use this component for the PR curve. However, the threshold values that the component iterates through do not include the value 1. It starts from 0 and iterates 100 times, with the final threshold value reaching 0.99.
As result PR-curve does start from (0,1).
So, my suggestion – change number of loops from 100 to 101.
I can copy the component and do it for myself, but it would be good for other users to have all points of threshold.
Thanks.
The component is really helpful.