best_movies.py

movie_names = movies.set_index('movieId')['title'].to_dict()
g=ratings.groupby('movieId')['rating'].count()

#TopMovies are top 3000 movies that got more user reviews than others. 
topMovies=g.sort_values(ascending=False).index.values[:3000]

uniq = ratings.movieId.unique()  # Ids of unique movies.
name2idx = {o:i for i,o in enumerate(uniq)}
topMovieIdx = np.array([name2idx[o] for o in topMovies])  # Indices of top rated movies in ratings.csv

get_movie_bias = models.Model(m_inp, m_b)  # traind movie biases for each movies in ratings.csv
movie_bias = get_movie_bias.predict(topMovieIdx) # lookup weights for top rated movies in movie biases embedding vector.

movie_ratings = [(b[0], movie_names[i]) for i,b in zip(topMovies,movie_bias)]  # --> [(movie bias, movie name), ...]
best_movies = sorted(movie_ratings, key=lambda o: o[0], reverse=True)[:48]  # Best 48 movies