Function gridsearch

pattern/vector/__init__.py:1935–1961

Returns the test results for every combination of optional parameters, using K-fold cross-validation for the given classifier (Bayes, KNN, SVM). For example: for (A, P, R, F), p in gridsearch(SVM, data, c=[0.1, 1, 10]): print (A, P, R, F), p > (0.919, 0.9

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

kfoldcv = K_fold_cv = k_fold_cv = k_fold_cross_validation = K_fold_cross_validation

def gridsearch(Classifier, documents=[], folds=10, **kwargs):
    """ Returns the test results for every combination of optional parameters,
        using K-fold cross-validation for the given classifier (Bayes, KNN, SVM).
        For example:
        for (A, P, R, F), p in gridsearch(SVM, data, c=[0.1, 1, 10]):
            print (A, P, R, F), p
        > (0.919, 0.921, 0.919, 0.920), {"c": 10}
        > (0.874, 0.884, 0.865, 0.874), {"c": 1}
        > (0.535, 0.424, 0.551, 0.454), {"c": 0.1}
    """
    def product(*args):
        # Yields the cartesian product of given iterables:
        # list(product([1, 2], [3, 4])) => [(1, 3), (1, 4), (2, 3), (2, 4)]
        p = [[]]
        for iterable in args:
            p = [x + [y] for x in p for y in iterable]
        for p in p:
            yield tuple(p)
    s = [] # [((A, P, R, F), parameters), ...]
    p = [] # [[("c", 0.1), ("c", 10), ...], 
           #  [("gamma", 0.1), ("gamma", 0.2), ...], ...]
    for k, v in kwargs.items():
        p.append([(k, v) for v in v])
    for p in product(*p):
        p = dict(p)
        s.append((K_fold_cross_validation(Classifier, documents, folds, **p), p))
    return sorted(s, reverse=True)

#--- NAIVE BAYES CLASSIFIER ------------------------------------------------------------------------