mlnext.score.ConfusionMatrix

class mlnext.score.ConfusionMatrix(TP: int = 0, FN: int = 0, TN: int = 0, FP: int = 0, DA: int = 0, TA: int = 0)

Bases: object

ConfusionMatrix is a confusion matrix for a binary classification problem. See https://en.wikipedia.org/wiki/Confusion_matrix.

Parameters:

• TP (int) – true positives, the number of samples from the positive class that are correctly assigned to the positive class.

• FN (int) – false negatives, the number of samples from the positive class that are wrongly assigned to the negative class.

• TN (int) – true negatives, the number of samples from the negative class that are correctly assigned to negative class.

• FP (int) – true negatives, the number of samples from the negative class that are wrongly assigned to the positive class.

• DA (int) – detected anomalies segments by at least one point.

• TA (int) – total number of anomaly segments.

Methods

metrics

Returns all metrics.

Attributes

DA
FN
FP
TA
TN
TP
accuracy	Calculates the accuracy (TP + TN) / (TP + TN + FP + FN).
f1	Calculates the F1-Score 2 * (precision * recall) / (precision + recall).
precision	Calculates the precision TP / (TP + FP).
recall	Calculates the recall TP / (TP + FN).
recall_anomalies	Calculates the percentage of detected anomaly segments.

__add__(cm: ConfusionMatrix) → ConfusionMatrix

Overrides the add operator.

Returns:

Returns a new matrix with feature-wise added values.

Return type:

ConfusionMatrix

property accuracy: float

Calculates the accuracy (TP + TN) / (TP + TN + FP + FN).

Returns:

Returns the accuracy.

Return type:

np.ndarray

property f1: float

Calculates the F1-Score 2 * (precision * recall) / (precision + recall).

Returns:

Returns the F1-score.

Return type:

np.ndarray

metrics() → Dict[str, float]

Returns all metrics.

Returns:

Returns an mapping of all performance metrics.

Return type:

T.Dict[str, float]

property precision: float

Calculates the precision TP / (TP + FP).

Returns:

Returns the precision.

Return type:

float

property recall: float

Calculates the recall TP / (TP + FN).

Returns:

Returns the recall.

Return type:

float

property recall_anomalies: float

Calculates the percentage of detected anomaly segments.

Returns:

Returns the percentage of detected segments.

Return type:

float