Test Items methodology

TestItemsMethodology(only_greater_eq=None)

Bases: Methodology

Class which, given a train set and a test set, has the task to calculate which items must be used in order to generate a recommendation list

With TestItemsMethodology, given a user \(u\), items to recommend for \(u\) are all items that appear in the test set of every user excluding those items that appear in the train set of \(u\)

If the only_greater_eq parameter is set, then only items with rating score \(>=\) only_greater_eq will be returned

PARAMETER	DESCRIPTION
only_greater_eq	float which acts as a filter, if specified only items with rating score \(>=\) only_greater_eq will be returned TYPE: float DEFAULT: None

Source code in clayrs/recsys/methodology.py

def __init__(self, only_greater_eq: float = None):
    super(TestItemsMethodology, self).__init__(only_greater_eq)

    self._filtered_test_set_items: Optional[np.ndarray] = None

filter_single(user_idx, train_set, test_set)

Method that returns items that need to be part of the recommendation list of a single user. Since it's the TestItems Methodology, all items that appear in the test set of every user will be returned, except for those that appear in the train set of the user passed as parameter

PARAMETER	DESCRIPTION
user_idx	User idx (meaning its mapped integer) of which we want to calculate items that must appear in its recommendation list TYPE: int
train_set	Ratings object which contains the train set of every user TYPE: Ratings
test_set	Ratings object which contains the test set of every user TYPE: Ratings

Source code in clayrs/recsys/methodology.py

def filter_single(self, user_idx: int, train_set: Ratings, test_set: Ratings) -> np.ndarray:
    """
    Method that returns items that need to be part of the recommendation list of a single user.
    Since it's the TestItems Methodology, all items that appear in the *test set* of every user will be returned,
    except for those that appear in the *train set* of the user passed as parameter

    Args:
        user_idx: User idx (meaning its mapped integer) of which we want to calculate items that must appear in its
            recommendation list
        train_set: `Ratings` object which contains the train set of every user
        test_set: `Ratings` object which contains the test set of every user
    """
    already_seen_items_it = pd.unique(train_set.get_user_interactions(user_idx)[:, 1].astype(int))

    self._query_vector[already_seen_items_it] = False

    result = self._items_arr[self._query_vector]

    self._query_vector[self._filtered_test_set_items] = True

    return result.astype(int)