Does the order of your columns matter when loading new data through existing XG boost model

#1

I have trained my model using 3 years of data, and then run new data through it but the probabilities seem infated compared to my existing model. The only difference I can see is the new data has the same features as the what the old model was trained on but the order of the columns in the dataframe is different, should this matter?

#2

Yes, the order of data columns matter. Please make sure that the columns of the test data match up with the columns of the training data, in exact same order.

1 Like