Skip to content

Recall

yohou.metrics.classification.Recall

Bases: BaseHardLabelScorer

Recall (sensitivity) from class-probability forecasts.

Computes recall by argmaxing predicted probabilities into hard labels and computing the ratio of true positives to all actual positives.

\[\text{Recall}_k = \frac{TP_k}{TP_k + FN_k}\]

Parameters

Name Type Description Default
average str

Class averaging strategy: "macro" (unweighted mean across classes), "micro" (aggregate counts first), or "weighted" (support-weighted mean).

 "macro"
zero_division float

Value returned when the denominator is zero.

 0.0
aggregation_method list of str or str

Dimensions to aggregate over.

 "all"
groups list of str, dict of str to float, or None

Panel group filter or filter with weights.

 None
components list of str, dict of str to float, or None

Component filter or filter with weights.

 None

Attributes

Name Type Description
lower_is_better bool

Always False (higher recall is better).

Examples

>>> import polars as pl
>>> from datetime import datetime
>>> from yohou.metrics.classification import Recall
>>> y_true = pl.DataFrame({
...     "time": [datetime(2020, 1, i) for i in range(1, 6)],
...     "weather": ["sunny", "rainy", "cloudy", "sunny", "rainy"],
... })
>>> y_pred = pl.DataFrame({
...     "vintage_time": [datetime(2019, 12, 31)] * 5,
...     "time": [datetime(2020, 1, i) for i in range(1, 6)],
...     "weather_proba_sunny": [0.7, 0.1, 0.2, 0.2, 0.1],
...     "weather_proba_rainy": [0.2, 0.8, 0.1, 0.1, 0.8],
...     "weather_proba_cloudy": [0.1, 0.1, 0.7, 0.7, 0.1],
... })
>>> scorer = Recall()
>>> _ = scorer.fit(y_true)
>>> scorer.score(y_true, y_pred)
0.833...

See Also

  • Precision : Precision (positive predictive value).
  • FBetaScore : Weighted harmonic mean of precision and recall.

Source Code

View on GitHub

Show/Hide source 
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
class Recall(BaseHardLabelScorer):
    r"""Recall (sensitivity) from class-probability forecasts.

    Computes recall by argmaxing predicted probabilities into hard labels
    and computing the ratio of true positives to all actual positives.

    $$\text{Recall}_k = \frac{TP_k}{TP_k + FN_k}$$

    Parameters
    ----------
    average : str, default="macro"
        Class averaging strategy: ``"macro"`` (unweighted mean across
        classes), ``"micro"`` (aggregate counts first), or ``"weighted"``
        (support-weighted mean).
    zero_division : float, default=0.0
        Value returned when the denominator is zero.
    aggregation_method : list of str or str, default="all"
        Dimensions to aggregate over.
    groups : list of str, dict of str to float, or None, default=None
        Panel group filter or filter with weights.
    components : list of str, dict of str to float, or None, default=None
        Component filter or filter with weights.

    Attributes
    ----------
    lower_is_better : bool
        Always False (higher recall is better).

    Examples
    --------
    >>> import polars as pl
    >>> from datetime import datetime
    >>> from yohou.metrics.classification import Recall
    >>> y_true = pl.DataFrame({
    ...     "time": [datetime(2020, 1, i) for i in range(1, 6)],
    ...     "weather": ["sunny", "rainy", "cloudy", "sunny", "rainy"],
    ... })
    >>> y_pred = pl.DataFrame({
    ...     "vintage_time": [datetime(2019, 12, 31)] * 5,
    ...     "time": [datetime(2020, 1, i) for i in range(1, 6)],
    ...     "weather_proba_sunny": [0.7, 0.1, 0.2, 0.2, 0.1],
    ...     "weather_proba_rainy": [0.2, 0.8, 0.1, 0.1, 0.8],
    ...     "weather_proba_cloudy": [0.1, 0.1, 0.7, 0.7, 0.1],
    ... })
    >>> scorer = Recall()
    >>> _ = scorer.fit(y_true)
    >>> scorer.score(y_true, y_pred)  # doctest: +ELLIPSIS
    0.833...

    See Also
    --------
    - [`Precision`][yohou.metrics.classification.Precision] : Precision (positive predictive value).
    - [`FBetaScore`][yohou.metrics.classification.FBetaScore] : Weighted harmonic mean of precision and recall.

    """

    _metric_name = "recall"
    _lower_is_better = False

    def _compute_metric_from_counts(self, counts: pl.DataFrame) -> pl.DataFrame:
        """Compute recall from confusion counts."""
        return counts.select(
            pl
            .when(pl.col("tp") + pl.col("fn") == 0)
            .then(self.zero_division)
            .otherwise(pl.col("tp") / (pl.col("tp") + pl.col("fn")))
            .alias("value")
        )