ML.NET Model Builder and Explainability (Permuation Feature Importance)
Hello,
I am trying to add PFI on the Trained model which I got through Model Builder.
But I think it's unable to unpack the trained model made by Model Builder.
Although I am trying to copy the training model with manually written code, by copying the trainer and its setting.
The performance (R2, RMS-loss) is always low in manual code.
I think model builder does something to make the best model, but it is hidden.
I am trying to use this method and write a research paper, but I feel the Model Builder is not replicable, and not transparent.
So goes back to the original question, is there a way to implement PFI to the trained model made by the model builder?
and can I manually train a model which is the same as the model builder made?