bpr-opt-06.py

#-----------------------
# FIND SIMILAR ARTISTS
#-----------------------

def find_similar_artists(artist=None, num_items=10):
    """Find artists similar to an artist.

    Args:
        artist (str): The name of the artist we want to find similar artists for

        num_items (int): How many similar artists we want to return.

    Returns:
        similar (pandas.DataFrame): DataFrame with num_items artist names and scores

    """

    # Grab our User matrix U
    user_vecs = get_variable(graph, session, 'user_factors')

    # Grab our Item matrix V
    item_vecs = get_variable(graph, session, 'item_factors')

    # Grab our item bias
    item_bi = get_variable(graph, session, 'item_bias').reshape(-1)

    # Get the item id for Lady GaGa
    item_id = int(item_lookup[item_lookup.artist == artist]['artist_id'])

    # Get the item vector for our item_id and transpose it.
    item_vec = item_vecs[item_id].T

    # Calculate the similarity between Lady GaGa and all other artists
    # by multiplying the item vector with our item_matrix
    scores = np.add(item_vecs.dot(item_vec), item_bi).reshape(1,-1)[0]

    # Get the indices for the top 10 scores
    top_10 = np.argsort(scores)[::-1][:num_items]

    # We then use our lookup table to grab the names of these indices
    # and add it along with its score to a pandas dataframe.
    artists, artist_scores = [], []
    
    for idx in top_10:
        artists.append(item_lookup.artist.loc[item_lookup.artist_id == str(idx)].iloc[0])
        artist_scores.append(scores[idx])

    similar = pd.DataFrame({'artist': artists, 'score': artist_scores})

    return similar

print(find_similar_artists(artist='beyoncé'))