Sign up to take part
Registered users can ask their own questions, contribute to discussions, and be part of the Community!
Registered users can ask their own questions, contribute to discussions, and be part of the Community!
I am wondering why there isn't an option to optimize a model's threshold based on custom code as is possible for the hyperparameter and model evaluation in Dataiku's AutoML. Is this a fundamental limitation that I'm not considering or is this a potential DSS feature request?
Thanks!
In my practice, using a cost matrix (and confirming lift with ROC) works well enough. What did you have in mind?
Blog that explains more: https://medium.com/better-ml/threshold-optimization-cost-a72903db36aa