Function confusion_matrix

pattern/metrics.py:101–123

Returns the performance of a binary classification task (i.e., predicts True or False) as a tuple of (TP, TN, FP, FN): - TP: true positives = correct hits, - TN: true negatives = correct rejections, - FP: false positives = false alarm (= type I error), -

(classify=lambda document: False, documents=[(None,False)])

MACRO = "macro"

def confusion_matrix(classify=lambda document: False, documents=[(None,False)]):
    """ Returns the performance of a binary classification task (i.e., predicts True or False)
        as a tuple of (TP, TN, FP, FN):
        - TP: true positives  = correct hits, 
        - TN: true negatives  = correct rejections,
        - FP: false positives = false alarm (= type I error), 
        - FN: false negatives = misses (= type II error).
        The given classify() function returns True or False for a document.
        The list of documents contains (document, bool)-tuples for testing,
        where True means a document that should be identified as True by classify().
    """
    TN = TP = FN = FP = 0
    for document, b1 in documents:
        b2 = classify(document)
        if b1 and b2:
            TP += 1 # true positive
        elif not b1 and not b2:
            TN += 1 # true negative
        elif not b1 and b2:
            FP += 1 # false positive (type I error)
        elif b1 and not b2:
            FN += 1 # false negative (type II error)
    return TP, TN, FP, FN

def test(classify=lambda document:False, documents=[], average=None):
    """ Returns an (accuracy, precision, recall, F1-score)-tuple.