def load_dermatology():
dermatology_file = pd.read_csv('datasets/dermatology.data', header=None)
dermatology_dataset = np.array(dermatology_file)
ages = [int(age) for age in dermatology_dataset[:, -2]]
dermatology_dataset[:, -2] = np.array(ages)
new_dermatology = dataset_utils.prepare_data(dermatology_dataset)
n_features = new_dermatology.shape[1] - 2
return new_dermatology, n_features
def load_verbetral():
vertebral_file = pd.read_csv('datasets/column_3C.dat',
header=None,
sep=' ')
vertebral_dataset = np.array(vertebral_file)
new_vertebral = dataset_utils.prepare_data(vertebral_dataset)
n_features = new_vertebral.shape[1] - 2
return new_vertebral, n_features
def load_breast_cancer():
breast_cancer_file = pd.read_csv('datasets/breast-cancer-wisconsin.csv',
header=None)
breast_cancer_dataset = np.array(breast_cancer_file)
labels_bc = breast_cancer_dataset[:, -1]
labels_bc = np.where(labels_bc == 4, 1, 0)
breast_cancer_dataset[:, -1] = labels_bc
new_breast_cancer = dataset_utils.prepare_data(breast_cancer_dataset)
n_features = new_breast_cancer.shape[1] - 2
return new_breast_cancer, n_features
def load_artificial_regression():
def regression_function(x):
return 3 * sin(x) + 1
N = 500
regression_dataset = np.random.normal(size=N).reshape((N, 1))
y_regression = np.array(
[regression_function(x) for x in regression_dataset]).reshape((N, 1))
regression_dataset = np.append(regression_dataset, y_regression, axis=1)
new_dataset = dataset_utils.prepare_data(np.array(regression_dataset))
n_features = 1
plt.scatter(new_dataset[:, 1], new_dataset[:, -1])
plt.show()
return new_dataset, n_features
def load_iris_dataset():
"""
Carrega o dataset, especificando qual classe
se deseja que seja representada pelo número 1.
As demais se tornam 0.
A label passada pode ser:
0 - setosa;
1 - versicolor;
2 - virginica;
"""
try:
iris = datasets.load_iris()
stream = open('configurations/irisConfigurations.yml',
'r',
encoding='utf-8').read()
iris_cfg = yaml.load(stream=stream, Loader=yaml.FullLoader)
n_features = len(iris_cfg['features'])
iris_classes = list(iris['target_names'])
label = iris_classes.index(iris_cfg['flower_to_classify'])
attributes = np.array(iris['data'])
if n_features < 4:
chosen_features = list(set(iris_cfg['features']))
features_indexes = sorted(
[iris['feature_names'].index(x) for x in chosen_features])
new_attributes = [attributes[:, i] for i in features_indexes]
attributes = np.array(new_attributes).T
labels = np.array(iris['target'])
labels = np.reshape(labels, [labels.shape[0], 1])
iris_dataset = np.append(attributes, labels, axis=1)
new_dataset = dataset_utils.transform_dataset(iris_dataset, label)
new_dataset = dataset_utils.prepare_data(new_dataset)
return new_dataset, n_features
except IndexError:
print('Erro ao carregar Iris.')
print('Verifique se o arquivo irisConfiguration.yml está correto.')
exit()
def load_multiclass_artificial():
n_features = 2
dataset = np.array([[np.random.uniform(1, 3), y, 0]
for y in np.random.uniform(4, 6, 50)])
plt.plot(dataset[:, 0], dataset[:, 1], 'ro')
dataset = np.append(dataset, [[np.random.uniform(4, 6), y, 1]
for y in np.random.uniform(1, 3, 50)],
axis=0)
plt.plot(dataset[50:-1, 0], dataset[50:-1, 1], 'y*')
dataset = np.append(dataset, [[np.random.uniform(7, 9), y, 2]
for y in np.random.uniform(4, 6, 50)],
axis=0)
plt.plot(dataset[100:-1, 0], dataset[100:-1, 1], 'b^')
plt.axis([0, 10, 0, 8])
plt.savefig('plots/artificial_data_plot.png')
plt.show()
new_dataset = dataset_utils.prepare_data(dataset)
return new_dataset, n_features
def artificial_data_p():
n_features = 2
dataset = np.array([[np.random.uniform(0, 0.5), y, 0]
for y in np.random.uniform(0, 0.5, 10)])
dataset = np.append(dataset, [[np.random.uniform(0, 0.5), y, 0]
for y in np.random.uniform(7, 7.5, 10)],
axis=0)
dataset = np.append(dataset, [[np.random.uniform(3, 3.5), y, 0]
for y in np.random.uniform(0, 0.5, 10)],
axis=0)
plt.plot(dataset[:, 0], dataset[:, 1], 'ro')
dataset = np.append(dataset, [[np.random.uniform(3, 3.5), y, 1]
for y in np.random.uniform(7, 7.5, 10)],
axis=0)
plt.plot(dataset[30:, 0], dataset[30:, 1], 'bo')
plt.axis([-1, 4, -1, 8])
plt.show()
new_dataset = dataset_utils.prepare_data(dataset)
return new_dataset, n_features
def load_multiclass_iris():
iris = datasets.load_iris()
stream = open('configurations/irisConfigurations.yml',
'r',
encoding='utf-8').read()
iris_cfg = yaml.load(stream=stream, Loader=yaml.FullLoader)
n_features = len(iris_cfg['features'])
attributes = np.array(iris['data'])
if n_features < 4:
chosen_features = list(set(iris_cfg['features']))
features_indexes = sorted(
[iris['feature_names'].index(x) for x in chosen_features])
new_attributes = [attributes[:, i] for i in features_indexes]
attributes = np.array(new_attributes).T
labels = np.array(iris['target'])
iris_dataset = np.append(attributes, labels[:, None], axis=1)
new_dataset = dataset_utils.prepare_data(iris_dataset)
return new_dataset, n_features