I have built a simple tree predictor and then applied this to a new set of data. This was giving a very different probability distribution so I tested connecting the same training data and comparing probability scores across the two predictions widgets in the following flow.The pre-processing steps are the same for both models (Group Mosaic + select columns), but the Predictions widgets give different outputs. Any idea why this might be? I'm on version 3.13.0.

Workflow

  • After a bit more investigation it appears that the 'Create Class' widget is the cause of the different scoring. If this is removed (along with the new feature it creates) then the probability scores from the two Prediction widgets do match. I'm not sure why this should be the case given the rules in the 'Create Class' widget are identical. Is it because by default the resulting class created is treated as a 'target' variable (note in the 'Select Columns' widget I've reset this to be a 'feature')? – Andrew Lockett May 31 at 8:50

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.