Default threshold while using Spline model for binary classification

[dnoted]

Hi There!

I am using the Spline model for binary classification (binary categorical response). I was just wondering if there's a way to understand if there's a default threshold that can be used for 0 and 1?

Thanks,

Divya

[JohnJPS · Accepted solution]

Hi @dnoted

I believe so.  When set for a binary response and run through a scoring process, the Spline model will generate a scores between 0 and 1.  From there, you can run that through a Formula tool and invoke any cut-off you desire.  I've done this in the attached workflow that runs against the well known Kaggle Titanic dataset.  If I adjust the formula that follows the Scoring node to something either above or below 0.5, my results just get worse.  I'm not sure, however, if that would always be the case. You could experiment on whatever dataset you're working with.

Hope that helps!

[dnoted]

Thanks, John. This is helpful! I can now use the ROC curve to pick a good threshold and interpret my output accordingly. I was wondering if Spline or other predictive models for that matter do this by default.