I have a binary classification problem, ran it through DAI, and am given test AUC's. Where do I find the probability threshold that is used during deployment to score new rows of data?
An example would be a threshold of .50; I.E a target variable >.50 gets a 1, and target<.50 gets a 0 (or vice versa) during a decision. I need the exact threshold beyond the 4 digit concatenated number that is shown in the GUI as you move across the AUC curve. In the pictures below I've matched the thresholds and can't get the same confusion matrices with the identical threshold. Notice the False Positives are very minimally different.
DAI Threshold @ .0301 Threshold
Sklearn Confusion Matrix @ .0301 Threshold
UPDATED ANSWER: Download the "Experiments Summary" tab after a completed experiment on DAI. Within the zip file you'll find an ensemble_roc_test.json that gives thresholds up to 10 digits.