In EDA we actually get insights about the data. That is completely fine but what i cannot understand is on which dataset should we do the EDA?
Should we do the EDA on train dataset or train + valid dataset or on the complete dataset.
I have seen many other questions similar to this and they have controversial answers. Question1 Question2 Should exploratory data analysis include validation set?
More precisely can anyone given an better explanation why i should/shouldn't use the validation dataset in the EDA than the above questions?.