move id issue

[teang1995]

Hi, can i get the script code for hasing movie id?
I want to get performance of NCF by domain(genre or popularity), so i have to get original id from your preprocessed data.
Thank you

[farouqzaib]

@teang1995
newMovieIdDict = dict(zip(ratings['movieId'].unique(), range(len(ratings['movieId'].unique()))))

^ is what I think they did. ratings here is the Dataframe from the original Movielens 1M data. The keys in the dictionary are the original movie IDs, and the values are their mapping to the ID used in the NCF paper. I've spot-checked a number of entries and the mapping holds true.

Perhaps someone from the NCF paper can confirm.