Closed Jensxy closed 5 years ago
Yes, of course. You just have to make sure, that you will provide a correct prediction function.
In the example here:
https://modeloriented.github.io/shapper/reference/individual_variable_effect.html
there is a random forest model, yet, you can use any model, just to parameter predict_function
pass a function that takes model and data and return predictions.
Is it possible to combine shapper with custom build models? I do not want to use a predefined algorithm like RandomForest.