def recommend(self, playlist_id, at=10):
playlist_id = int(playlist_id)
helper = Helper()
### DUE TO TIME CONSTRAINT THE CODE STRUCTURE HERE IS REDUNTANT
### TODO exploit inheritance to reduce code duplications and simple extract ratings, combine them, simply by iterate over a list of recommenders
### COMMON CODE ###
self.hybrid_ratings = None #BE CAREFUL, MAGIC INSIDE :)
### COMBINE RATINGS IN DIFFERENT WAYS (seq, random short, random long)
if (helper.is_sequential(playlist_id)):
self.userCF_ratings = self.userCF_sequential.get_expected_ratings(
playlist_id)
self.itemCF_ratings = self.itemCF_sequential.get_expected_ratings(
playlist_id)
self.cbf_ratings = self.cbf_sequential.get_expected_ratings(
playlist_id)
self.slim_elastic_ratings = self.slim_elastic_sequential.get_expected_ratings(
playlist_id)
# self.svd_icm_ratings = self.svd_icm_sequential.get_expected_ratings(playlist_id)
self.ALS_ratings = self.ALS_sequential.get_expected_ratings(
playlist_id)
self.slim_ratings = self.slim_sequential.get_expected_ratings(
playlist_id)
w_right = self.w_seq
else:
self.userCF_ratings = self.userCF.get_expected_ratings(playlist_id)
self.itemCF_ratings = self.itemCF.get_expected_ratings(playlist_id)
self.cbf_ratings = self.cbf.get_expected_ratings(playlist_id)
self.slim_elastic_ratings = self.slim_elastic.get_expected_ratings(
playlist_id)
# self.svd_icm_ratings = self.svd_icm.get_expected_ratings(playlist_id)
self.ALS_ratings = self.ALS.get_expected_ratings(playlist_id)
self.slim_ratings = self.slim_random.get_expected_ratings(
playlist_id)
if len(self.URM[playlist_id].indices) > 10:
w_right = self.w_long
else:
w_right = self.w_short
self.hybrid_ratings = self.userCF_ratings * w_right["user_cf"]
self.hybrid_ratings += self.itemCF_ratings * w_right["item_cf"]
self.hybrid_ratings += self.cbf_ratings * w_right["cbf"]
self.hybrid_ratings += self.slim_ratings * w_right["slim"]
# self.hybrid_ratings += self.svd_icm_ratings * w_right["svd_icm"]
self.hybrid_ratings += self.ALS_ratings * w_right["als"]
self.hybrid_ratings += self.slim_elastic_ratings * w_right["elastic"]
recommended_items = np.flip(np.argsort(self.hybrid_ratings), 0)
# REMOVING SEEN
unseen_items_mask = np.in1d(recommended_items,
self.URM[playlist_id].indices,
assume_unique=True,
invert=True)
recommended_items = recommended_items[unseen_items_mask]
return recommended_items[0:at]
