def predict(self,
to_predict,
targets,
top_n=5,
remove_known=True,
mask=True):
targets = targets['track_id'].unique()
test = to_predict['playlist_id'].unique()
predictions = []
for i in tqdm(test):
playlist_index = utils.get_target_index(i, self.playlists)
current_ratings = self.recommender.urm[playlist_index, :]
ratings = self.recommender.predict(current_ratings, remove_known)
if mask:
not_selected = np.where(~np.in1d(self.tracks, targets))[0]
ratings[not_selected] = 0
top_n_indexes = utils.get_n_best_indexes(ratings, top_n)
top_n_predictions = self.tracks[top_n_indexes]
predictions.append((i, top_n_predictions))
self.predictions = predictions
