"I understand that the feature weights in the AI predictor never change unless the predictor is retrained.
However, is there a way to find out:
- Which features are included in the predictor
- The number of features in the predictor
- The weight assigned to each individual feature in the predictor
This information would be incredibly interesting and provide much deeper insights into the 'black box' of the ML model. It also offers some transfer value to the current ranking system.
I understand that to some extent, I can find some information under feature stats, specifically 'Target Information Regression, but it does not say anything what the final predictor actually contain.