Hey everybody ,
i trained a model for tabular based off the fastai doc’s.
its give me a decent accuracy of about 89% but when i submit it to kaggle i get a score of 0.58 which is way off
first i thought it was overfitting, but after changing the hyperparameters i realized it might be more than that
Has anyone had a similar experience or does anybody have any ideas on that?Any help will be greatly appreciated.