Function K_fold_cross_validation

pattern/vector/__init__.py:1911–1931

For 10-fold cross-validation, performs 10 separate tests of the classifier, each with a different 9/10 training and 1/10 testing documents. The given classifier is a class (Bayes, KNN, SVM) which is initialized with the given optional parameters.

(Classifier, documents=[], folds=10, **kwargs)

        return s

def K_fold_cross_validation(Classifier, documents=[], folds=10, **kwargs):
    """ For 10-fold cross-validation, performs 10 separate tests of the classifier,
        each with a different 9/10 training and 1/10 testing documents.
        The given classifier is a class (Bayes, KNN, SVM)
        which is initialized with the given optional parameters.
    """
    K = kwargs.pop("K", folds)
    d = [isinstance(d, Document) and (d, d.type) or d for d in documents]
    d = shuffled(d) # Avoid a list sorted by type (because we take successive folds).
    m = [0.0, 0.0, 0.0, 0.0] # Macro-average accuracy, precision, recall & F1-score.
    for i in range(K):
        n = len(d) / float(K)     # Test fold size.
        x = int(round(i * n))     # Test fold start index.
        y = int(round(i * n + n)) # Test fold stop index.
        classifier = Classifier(train=d[:x]+d[y:], **kwargs)
        A, P, R, F = classifier.test(d[x:y], **kwargs)
        m[0] += A
        m[1] += P
        m[2] += R
        m[3] += F
    return tuple([v / (K or 1) for v in m])
    
kfoldcv = K_fold_cv = k_fold_cv = k_fold_cross_validation = K_fold_cross_validation