Skip to content

dcg

logger = logging.getLogger(__name__) module-attribute

DCGK

Bases: ListwiseMetricK

Computes the sum of gains of all items in a recommendation list.

Relevant items that are ranked higher in the Top-K recommendations have a higher gain.

The Discounted Cumulative Gain (DCG) is computed for every user as

.. math::

\text{DiscountedCumulativeGain}(u) = \sum\limits_{i \in Top-K(u)} \frac{y^{true}_{u,i}}{\log_2 (\text{rank}(u,i) + 1)}

:param K: Size of the recommendation list consisting of the Top-K item predictions. :type K: int

This code is adapted from RecPack :cite:recpack

Source code in src/streamsight/metrics/ranking/dcg.py 
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
class DCGK(ListwiseMetricK):
    """Computes the sum of gains of all items in a recommendation list.

    Relevant items that are ranked higher in the Top-K recommendations have a higher gain.

    The Discounted Cumulative Gain (DCG) is computed for every user as

    .. math::

        \\text{DiscountedCumulativeGain}(u) = \\sum\\limits_{i \\in Top-K(u)} \\frac{y^{true}_{u,i}}{\\log_2 (\\text{rank}(u,i) + 1)}


    :param K: Size of the recommendation list consisting of the Top-K item predictions.
    :type K: int

    This code is adapted from RecPack :cite:`recpack`
    """
    IS_BASE: bool = False

    def _calculate(self, y_true: csr_matrix, y_pred: csr_matrix) -> None:
        logger.debug("Precision compute started - %s", self.name)
        logger.debug("Shape of matrix: (%d, %d)", y_true.shape[0], y_true.shape[1])
        logger.debug("Number of ground truth interactions: %d", y_true.nnz)

        denominator = y_pred.multiply(y_true)
        # Denominator: log2(rank_i + 1)
        denominator.data = np.log2(denominator.data + 1)
        # Binary relevance
        # Numerator: rel_i
        numerator = y_true

        dcg = sparse_divide_nonzero(numerator, denominator)

        self._scores = csr_matrix(dcg.sum(axis=1))

        logger.debug(f"DCGK compute complete - {self.name}")

IS_BASE = False class-attribute instance-attribute

name property

Name of the metric.

params property

Parameters of the metric.

identifier property

Identifier of the object.

Identifier is made by combining the class name with the parameters passed at construction time.

Constructed by recreating the initialisation call. Example: Algorithm(param_1=value)

:return: Identifier of the object

micro_result property

User level results for the metric.

Contains an entry for every user.

:return: The results DataFrame with columns: user_id, score :rtype: pd.DataFrame

macro_result property

Global metric value obtained by taking the average over all users.

:raises ValueError: If the metric has not been calculated yet. :return: The global metric value. :rtype: float, optional

is_time_aware property

Whether the metric is time-aware.

timestamp_limit property

The timestamp limit for the metric.

num_items property

Dimension of the item-space in both y_true and y_pred

num_users property

Dimension of the user-space in both y_true and y_pred after elimination of users without interactions in y_true.

K = K instance-attribute

col_names property

The names of the columns in the results DataFrame.

get_params()

Get the parameters of the metric.

Source code in src/streamsight/metrics/core/base.py 
53
54
55
56
57
def get_params(self) -> dict[str, int | None]:
    """Get the parameters of the metric."""
    if not self.is_time_aware:
        return {}
    return {"timestamp_limit": self._timestamp_limit}

calculate(y_true, y_pred)

Calculates this metric for all nonzero users in y_true, given true labels and predicted scores.

Source code in src/streamsight/metrics/core/base.py 
116
117
118
119
120
121
def calculate(self, y_true: csr_matrix, y_pred: csr_matrix) -> None:
    """Calculates this metric for all nonzero users in `y_true`,
    given true labels and predicted scores.
    """
    y_true, y_pred = self._prepare_matrix(y_true, y_pred)
    self._calculate(y_true, y_pred)

prepare_matrix(y_true, y_pred)

Source code in src/streamsight/metrics/core/top_k.py 
57
58
59
60
def prepare_matrix(self, y_true: csr_matrix, y_pred: csr_matrix) -> tuple[csr_matrix, csr_matrix]:
    y_true, y_pred = super()._prepare_matrix(y_true, y_pred)
    y_pred = get_top_K_ranks(y_pred, self.K)
    return y_true, y_pred