I observe we are applying encoder.transform on training and validation set in the lesson notebook.
I believe this is not correct. On training set we should apply only
endocer.fit_transfrom and on the validation and test sets we should apply only
Now the next question can be what is the difference btw fit_transform and transform. I recently came across an article explaining this very clearly. This small change can go a long way is resolving the overfitting on the test sets.
Please do share your thought if u think otherwise.