Probability of prediction column in a binary classification problem.

SA_H

Using the (apply model operator) in the binary classification problem, the results exported in the form of the prediction column and confidences of being 0 or 1. Please I am wondering if there is an operator that can export the probability (range from 0-1 such as 0, 0.1, 0.2, ...) or scores indicating the likelihood that a label comes from a particular class (0 or 1).

varunm1

Hello@summer_helmi

Sorry, I am confused can you please explain in detail. The prediction confidences are similar to probabilities, that shows the confidences of a particular class for a particular observation between 0 and 1. If the algorithm is really confident then it will assign 1 to related label (class).

MartinLiebig

Hi,
confidences are not probabilities, but correlated. To get something way closer to propabilities we recommend to use Rescale Confidences (Logistic), which uses Platt Scaling.
Best,
Martin

SA_H

Thank you,