Hello,
It looks like the XGBoost model converted by CoreML returns completely wrong outputs.
So, I do the following:
1) Create an xgboost model in python 3.5
2) Train the xgboost classifier in 4 categories.
3) Load the xgboost classifier in CoreML
4) Compare predictions between the two
So, whereas the output of the xgboost model is 'A', 'B', 'C', 'D', coreml outputs real numbers between -1 and 4 or 5.
I am using XGboost version 0.7.
Thank you