def accuracy_for_iris():
iris = load_iris()
folds = 3
data_split, labels_split = tools.cross_validation_split(dataset=iris.data,
labels=iris.target,
folds=folds)
print("Multinomial:")
accuracy_of_multinomial(data_split, labels_split)
print("Gaussian:")
accuracy_of_gaussian(data_split, labels_split)
def accuracy_of_multinomial(dataset, labels, num_of_bins=100):
folds = 3
data_split, labels_split = tools.cross_validation_split(dataset=dataset,
labels=labels,
folds=folds)
nb = NaiveBayesMultinomial(num_of_bins=num_of_bins)
clf = MultinomialNB(alpha=0.0001, fit_prior=True)
accuracies, sklearn_accuracies = tools.accuracy_of_method(
data_split, labels_split, nb, sklearn_class=clf)
return np.mean(accuracies), np.mean(sklearn_accuracies)
def accuracy_for_letters():
with open('../../data_sources/letter-recognition.data', 'r') as f:
dataset, labels = tools.load_text_file(f,
label_index=0,
labels_numeric=False)
folds = 3
data_split, labels_split = tools.cross_validation_split(
dataset=dataset, labels=labels, folds=folds)
# iris = load_iris()
print("Multinomial:")
accuracy_of_multinomial(data_split, labels_split, num_of_bins=11)
print("Gaussian:")
accuracy_of_gaussian(data_split, labels_split)
def accuracy_for_cancer():
with open('../../data_sources/kag_risk_factors_cervical_cancer.csv',
'r') as f:
dataset, labels = tools.load_text_file(f, label_index=28, dtype=float)
# labels = np.array(dataset[:,28], dtype=int)
# dataset=np.append(dataset[:,:28], dataset[:, 29:], axis=1)
# iris = load_iris()
# labels -= 1
folds = 3
data_split, labels_split = tools.cross_validation_split(
dataset=dataset, labels=labels, folds=folds)
# iris = load_iris()
print("Multinomial:")
accuracy_of_multinomial(data_split, labels_split)
print("Gaussian:")
accuracy_of_gaussian(data_split, labels_split)
def accuracy_for_trees():
with open('../../data_sources/covtype.csv', 'r') as f:
dataset, labels = tools.load_text_file(f,
label_index=-1,
dtype=float,
labels_numeric=True)
# iris = load_iris()
# labels -= 1
folds = 3
data_split, labels_split = tools.cross_validation_split(
dataset=dataset, labels=labels, folds=folds)
# iris = load_iris()
print("Multinomial:")
accuracy_of_multinomial(data_split, labels_split)
print("Gaussian:")
accuracy_of_gaussian(data_split, labels_split)
def accuracies_of_different_methods(self, dataset, labels, algorithm_classes, algorithm_kwargs):
"""
Returns np array of accuracies of different methods for one dataset, dataset cannot be already splitted, labels have to be separated
algorithm_classes - not_initiated class
algorithm_kwargs - arguments for initiation
:param dataset:
:param labels:
:param algorithm_classes:
:param algorithm_kwargs:
:return:
"""
data_split, labels_split = tools.cross_validation_split(dataset=abs(dataset), labels=labels, folds=3)
accuracies_for_dataset = np.zeros(len(algorithm_classes))
for index, algorithm in enumerate(algorithm_classes):
alg = algorithm(**algorithm_kwargs[index])
acc_my, _ = tools.accuracy_of_method(data_split, labels_split, alg)
accuracies_for_dataset[index] = np.mean(acc_my)
return accuracies_for_dataset