Merge pull request #140 from ita9naiwa/ita/add-filtering-already-liked

Add an option to whether to include previously liked items or not
benfred · Jul 25, 2018 · f5a9a56 · f5a9a56
2 parents c0b63cb + c61e2f7
commit f5a9a56
Show file tree

Hide file tree

Showing 2 changed files with 12 additions and 5 deletions.
diff --git a/implicit/nearest_neighbours.py b/implicit/nearest_neighbours.py
@@ -23,15 +23,17 @@ def fit(self, weighted):
         self.similarity = all_pairs_knn(weighted, self.K, show_progress=self.show_progress).tocsr()
         self.scorer = NearestNeighboursScorer(self.similarity)
 
-    def recommend(self, userid, user_items, N=10, filter_items=None, recalculate_user=False):
+    def recommend(self, userid, user_items,
+                  N=10, filter_already_liked_items=True, filter_items=None, recalculate_user=False):
         """ returns the best N recommendations for a user given its id"""
         # recalculate_user is ignored because this is not a model based algorithm
         items = N
         if filter_items:
             items += len(filter_items)
 
         indices, data = self.scorer.recommend(userid, user_items.indptr, user_items.indices,
-                                              user_items.data, K=items)
+                                              user_items.data, K=items,
+                                              remove_own_likes=filter_already_liked_items)
         best = sorted(zip(indices, data), key=lambda x: -x[1])
 
         if not filter_items:

diff --git a/implicit/recommender_base.py b/implicit/recommender_base.py
@@ -25,7 +25,8 @@ def fit(self, item_users):
         pass
 
     @abstractmethod
-    def recommend(self, userid, user_items, N=10, filter_items=None, recalculate_user=False):
+    def recommend(self, userid, user_items,
+                  N=10, filter_already_liked_items=True, filter_items=None, recalculate_user=False):
         """
         Recommends items for a user
 
@@ -120,11 +121,15 @@ def __init__(self):
         # cache of item norms (useful for calculating similar items)
         self._item_norms = None
 
-    def recommend(self, userid, user_items, N=10, filter_items=None, recalculate_user=False):
+    def recommend(self, userid, user_items,
+                  N=10, filter_already_liked_items=True, filter_items=None, recalculate_user=False):
         user = self._user_factor(userid, user_items, recalculate_user)
 
         # calculate the top N items, removing the users own liked items from the results
-        liked = set(user_items[userid].indices)
+        if filter_already_liked_items is True:
+            liked = set(user_items[userid].indices)
+        else:
+            liked = set()
         scores = self.item_factors.dot(user)
         if filter_items:
             liked.update(filter_items)