from sklearn.preprocessing import MultiLabelBinarizer

popular_genres = ['Comedy',
 'Action',
 'Fantasy',
 'Adventure',
 'Kids',
 'Drama',
 'Sci-Fi',
 'Music',
 'Shounen',
 'Slice of Life']

def create_genre_flags(df, popular_genres):
    df = df.dropna(subset=['Genres'])
    df['Genres'] = df['Genres'].apply(lambda x:",".join(s.strip() for s in x.split(",")))
    # use MultiLabelBinarizer to create a one-hot encoded dataframe of the genres
    mlb = MultiLabelBinarizer()
    genre_df = pd.DataFrame(mlb.fit_transform(df['Genres'].str.split(',')),
                            columns=mlb.classes_,
                            index=df.index)
    # create a new dataframe with the movie id and genre columns
    new_df = pd.concat([df['anime_id'], genre_df[popular_genres]], axis=1)
    new_df.columns = ['anime_id'] + popular_genres
    return new_df

anime_genre_info_df = create_genre_flags(anime_info_df,popular_genres)
anime_info_df_final = anime_info_df.merge(anime_genre_info_df,on='anime_id')
anime_info_df_final.columns = [col if col=='anime_id' else f"ANIME_FEATURE {col}".upper() for col in anime_info_df_final.columns]
user_info.columns = [col if col=='user_id' else f"USER_FEATURE {col}".upper() for col in user_info.columns]

train_interim = relavence_scores.merge(anime_info_df_final)
train = train_interim.merge(user_info,how='inner')