mlnext.score.PRCurve#

class mlnext.score.PRCurve(tps: ndarray, fns: ndarray, tns: ndarray, fps: ndarray, das: ndarray, tas: ndarray, precision: ndarray, recall: ndarray, thresholds: ndarray)[source]#

Bases: object

Container for the result of pr_curve. Additionally computes the F1-score for each threshold. Can be indexed and returns a ConfusionMatrix for the i-th threshold.

Parameters:

tps (np.ndarray) – An increasing count of true positives, at index i being the number of positive samples assigned a score >= thresholds[i].
fns (np.ndarray) – A count of false negatives, at index i being the number of positive samples assigned a score < thresholds[i].
tns (np.ndarray) – A count of true negatives, at index i being the number of negative samples assigned a score < thresholds[i].
fps (np.ndarray) – A count of false positives, at index i being the number of negative samples assigned a score >= thresholds[i].
das (np.ndarray) – A count of detected anomaly segments, at index i being the number of detected anomalies for a score >= thresholds[i].
tas (np.ndarray) – Total number of anomaly segments.
precision (np.ndarray) – Precision values such that element i is the precision of predictions with score >= thresholds[i].
recall (np.ndarray) – Decreasing recall values such that element i is the recall of predictions with score >= thresholds[i].
thresholds (np.ndarray) – Increasing thresholds on the decision function used to compute precision and recall.

Methods

to_tensorboard

Converts the container to keyword arguments for Tensorboard.

Attributes

`accuracy`	Calculates the accuracy where at index i, the accuracy is the percentage of correctly assigned samples.
`auc`	Calculates the area-under-curve (auc).
`f1`	Calculates the F1-score.
`recall_anomalies`	Calculates the fraction of detected anomaly segments.
`tps`
`fns`
`tns`
`fps`
`das`
`tas`
`precision`
`recall`
`thresholds`

property accuracy: ndarray#

Calculates the accuracy where at index i, the accuracy is the percentage of correctly assigned samples.

Returns:: Returns the accuracy.
Return type:: np.ndarray

property auc: float#

Calculates the area-under-curve (auc).

Returns:: Returns the area-under-curve (auc) for the precision-recall curve.
Return type:: float

property f1: ndarray#

Calculates the F1-score.

Returns:: Returns the F1-score.
Return type:: np.ndarray

property recall_anomalies: ndarray#

Calculates the fraction of detected anomaly segments.

Returns:: Returns the anomaly recall.
Return type:: np.ndarray

to_tensorboard() → Dict[str, Any][source]#

Converts the container to keyword arguments for Tensorboard. See https://github.com/tensorflow/tensorboard/blob/master/tensorboard/plugins/pr_curve/README.md.

Returns:: Returns the pr-curve format expected for Tensorboard.
Return type:: T.Dict[str, T.Any]