This module contains the:

  • MetricNameEnum: Available metric names.

  • MetricScopeEnum: Metric scopes (micro, macro, class).

  • MetricTypeEnum: The type of data the metric operates on (probabilities, labels, or scores).

The metrics and their scopes are separated to allow flexible combinations, and easily identify the metric name and scope individually in the codebase.

MetricNameEnum

Bases: Enum

Available metric function names.

Source code in metrics_toolbox/metrics/enums.py 
17
18
19
20
21
22
23
24
25
26
class MetricNameEnum(Enum):
    """Available metric function names."""

    ROC_AUC = "roc_auc"
    PRECISION = "precision"
    ACCURACY = "accuracy"
    RECALL = "recall"
    F1_SCORE = "f1_score"
    MSE = "mse"
    RMSE = "rmse"

MetricScopeEnum

Bases: Enum

Available metric scopes. (micro, macro, target)

Attributes:
  • MICRO (str) –

    Micro-averaged metric scope

  • MACRO (str) –

    Macro-averaged metric scope

  • TARGET (str) –

    Class/Column/Signal scope. This is a generic term for "per class" metrics for classification and "per column" metrics for regression. Also the binary classification metrics with two classes fall into this category, and the target name would be the positive class.
Source code in metrics_toolbox/metrics/enums.py 
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
class MetricScopeEnum(Enum):
    """Available metric scopes. (micro, macro, target)

    Attributes
    ----------
    MICRO : str
        Micro-averaged metric scope
    MACRO : str
        Macro-averaged metric scope
    TARGET : str
        Class/Column/Signal scope. This is a generic term for "per class" metrics
        for classification and "per column" metrics for regression.
        Also the binary classification metrics with two classes fall into this category,
        and the target name would be the positive class.
    """

    MICRO = "micro"
    MACRO = "macro"
    TARGET = "target"

MetricTypeEnum

Bases: Enum

Available metric types. (probs vs labels, etc.)

Attributes:
  • PROBS (str) –

    Metric operates on predicted probabilities 0-1.

  • LABELS (str) –

    Metric operates on predicted class labels, string or int.

  • SCORES (str) –

    Metric operates on predicted scores, e.g. regression predictions.
Source code in metrics_toolbox/metrics/enums.py 
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
class MetricTypeEnum(Enum):
    """Available metric types. (probs vs labels, etc.)

    Attributes
    ----------
    PROBS : str
        Metric operates on predicted probabilities 0-1.
    LABELS : str
        Metric operates on predicted class labels, string or int.
    SCORES : str
        Metric operates on predicted scores, e.g. regression predictions.
    """

    PROBS = "probs"
    LABELS = "labels"
    SCORES = "scores"