learn module

learn.auc(x, y)

Computes the area under curve of given ROC-values by applying the trapezodial rule.

Parameters

xarray

x-values of the given funktion e.g. the FPR-values.

yarray

y-values of the given funktion e.g. the TPR-values.

Returns

float

The area under curve.

learn.confusion_matrix(y_true, y_pred, labels=None)

Creates a confusion matrix from the binary classification and ground truth.

Parameters

y_truearray

The ground truth labels.

y_predarray

The predicted labels.

labelstupel, optional

The desired labels. The default is None.

Returns

matrixndarray

A 2x2 matrix containing the True Positive, True Negatife, False Positive and False Negative.

learn.precision_recall_curve(y_true, probas_pred)

Computes the precision-recall-curve for given predicions. Much like the roc-funktion it takes two arrays of the same length. One with the predicting scores and one with the ground truth.

Parameters

y_trueTYPE

The array with the ground truth in binary.

probas_predTYPE

The array with the predicting scores.

Returns

precisionsarray

An array with the precisions for each threshold.

recallsarray

An array with the recall or sensitivity for each threshold.

thresholdsarray

An array with the tresholds.

learn.roc_curve(y_true, y_score)

Computes the ROC-values for given scores. It takes two arrays. One with the ground truth in binary an one with the scores. Both arrays must be of the same length, otherwise an error occurs. Example:

# generating dummy ground truth
a1 = np.zeros(10, dtype=int)
a2 = np.ones(10, dtype=int)
dummy_ground_truth = np.concatenate((a1, a2))
np.random.shuffle(dummy_ground_truth)

dummy_scores = np.random.rand(20)

fpr_dummy, tpr_dummy, thresh_dummy = roc_curve(dummy_ground_truth, dummy_scores)

print('FPR:', fpr_dummy)
print('TPR:', tpr_dummy)
print('thresholds:' thresh_dummy)

Parameters

y_truearray

The array with the graund truth values e.g. the annotation.

y_scorearray

The array with the scoring values.

Returns

fprarray

The false positive rates for every threshold.

tprarray

The true positive rate for every threshold.

thresholdsarray

The thresholds.