Metrics

class utils.metrics.Metrics(y_true: ndarray, y_pred: ndarray, classes: None | ndarray = None)

Bases: object

__init__(y_true: ndarray, y_pred: ndarray, classes: None | ndarray = None)

A collection of metrics that evaluate the results of classification.

Parameters:

• y_true (np.ndarray) – The actual class vector/matrix (assuming one-hot encoding in the latter case).

• y_pred (np.ndarray) – The predicted class vector/matrix (assuming one-hot encoding in the latter case).

• classes (None | np.ndarray) – Vector of unique class values. Required for single output classification accuracy calculation.

Variables:

_metric_modes – Map for defining which metric is suitable for a certain target type.

Returns:

None

Return type:

None

accuracy_oh() → float

Accuracy with one-hot encoded target. :return: Classification accuracy :rtype: float

accuracy_reg() → float

Accuracy of classification calculated from single output (regression). :return: Classification accuracy :rtype: float

calculate(metrics: list[Callable[[], float]] | str = 'all') → dict[str, float]

Calculate the given metrics based on the target labels. The functions are called based on the class vector shape.

Parameters:

metrics – List of functions to call. Alternatively, “all” to calculate all metrics. Default value is “all”.

Returns:

Dictionary with metric names as keys and scores as values.

Return type:

dict[str, float]

root_mean_squared() → float

Root-Mean Squared error. :return: RMS error :rtype: float

root_mean_squared_oh() → float

Root-Mean Squared error with one-hot encoded target. :return: RMS error :rtype: float

scaled_root_mean_squared() → float

Scaled Root-Mean Squared error. Used with target vectors scaled between 0.1 and 0.9. :return: SRMS error :rtype: float

scaled_root_mean_squared_oh() → float

Scaled Root-Mean Squared error with one-hot encoded target. Used with target vectors scaled between 0.1 and 0.9. :return: SRMS error :rtype: float