def setUp(self):
# classification problem.
iris = pd.read_csv("iris.csv")
self.iris = Dataset(iris)
# Regression problem
tips = pd.read_csv("tips.csv")
self.tips = Dataset(tips, dependent_col="tip")
def setUp(self):
# classification problem.
iris = pd.read_csv('iris.csv')
self.iris = Dataset(iris)
# Regression problem
tips = pd.read_csv('tips.csv')
self.tips = Dataset(tips, dependent_col='tip')
def generate_metafeatures_from_server(file_id, target_field, **kwargs):
# Read the data set into memory
raw_data = get_file_from_server(file_id)
df = pd.read_csv(StringIO(raw_data), sep=None, engine='python', **kwargs)
dataset = Dataset(df,
dependent_col=target_field,
prediction_type='classification')
return generate_metafeatures(dataset, target_field)
def get_metafeatures(df):
dataset = Dataset(df, dependent_col = 'class', prediction_type='classification')
meta_features = OrderedDict()
for i in dir(dataset):
result = getattr(dataset, i)
if not i.startswith('__') and not i.startswith('_') and hasattr(result, '__call__'):
meta_features[i] = result()
return meta_features
def generate_metafeatures_from_filepath(input_file, target_field, **kwargs):
"""Calls metafeature generating methods from dataset_describe"""
# Read the data set into memory
df = pd.read_csv(input_file, sep=None, engine='python', **kwargs)
dataset = Dataset(df,
dependent_col=target_field,
prediction_type='classification')
return generate_metafeatures(dataset, target_field)
class Dataset_Describe(unittest.TestCase):
def setUp(self):
# classification problem.
iris = pd.read_csv('iris.csv')
self.iris = Dataset(iris)
# Regression problem
tips = pd.read_csv('tips.csv')
self.tips = Dataset(tips, dependent_col='tip')
def test_number_of_rows(self):
result = self.iris.n_rows()
self.assertEqual(150, result)
def test_number_of_columns(self):
result = self.iris.n_columns()
self.assertEqual(5, result)
def test_if_self_categorical_cols_is_zero_iris(self):
result = self.iris.categorical_cols
self.assertEqual([], result)
def test_number_of_categorical_vars(self):
result = self.iris.n_categorical()
self.assertEqual(0, result)
def test_number_of_numerical_vars(self):
result = self.iris.n_numerical()
self.assertEqual(4, result)
def test_total_nclasses(self):
result = self.iris.n_classes()
self.assertEqual(3, result)
def test_total_nclasses_in_regression_problem(self):
result = self.tips.n_classes()
self.assertTrue(math.isnan(result))
def test_prediction_type_classification(self):
result = self.iris.prediction_type
self.assertEqual('classification', result)
def test_prediction_type_regression(self):
result = self.tips.prediction_type
self.assertEqual('regression', result)
def test_max_corr_with_dependent_classification(self):
result = self.iris.corr_with_dependent_abs_max()
self.assertTrue(math.isnan(result))
def test_max_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_max()
self.assertAlmostEqual(0.675, result, places=2)
def test_min_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_min()
self.assertAlmostEqual(0.002, result, places=2)
def test_25p_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_25p()
self.assertAlmostEqual(0.055, result, places=2)
def test_mean_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_mean()
self.assertAlmostEqual(0.18, result, places=2)
def test_median_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_median()
self.assertAlmostEqual(0.095, result, places=2)
def test_75p_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_75p()
self.assertAlmostEqual(0.125, result, places=2)
def test_std_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_std()
self.assertAlmostEqual(0.234, result, places=2)
def test_class_probablity_total_suite(self):
## evaluating for regression problem.
result = self.tips.class_prob_min()
self.assertTrue(math.isnan(result))
## evaluating for iris dataset.
## not really a useful test it seems, should get another dataset.
## Also not a unit test.
self.assertAlmostEqual(0.33, self.iris.class_prob_min(), places=2)
self.assertAlmostEqual(0.33, self.iris.class_prob_max(), places=2)
self.assertAlmostEqual(0, self.iris.class_prob_std(), places=2)
self.assertAlmostEqual(0.33, self.iris.class_prob_mean(), places=2)
self.assertAlmostEqual(0.33, self.iris.class_prob_median(), places=2)
def test_class_symbols_suite(self):
mean = 2.5
std = 1
min_ = 2
max_ = 4
## Iris has no nominal variables.
result = self.iris.symbols_sum()
self.assertTrue(math.isnan(result))
## evaluating for tips dataset.
## Also not a unit test.
symbol_counts = self.tips._get_symbols_per_category()
self.assertEqual(symbol_counts['sex'], 2)
self.assertEqual(symbol_counts['smoker'], 2)
self.assertAlmostEqual(max_, self.tips.symbols_max(), places=2)
self.assertAlmostEqual(min_, self.tips.symbols_min(), places=2)
self.assertAlmostEqual(std, self.tips.symbols_std(), places=2)
self.assertAlmostEqual(mean, self.tips.symbols_mean(), places=2)
class Dataset_Describe(unittest.TestCase):
def setUp(self):
# classification problem.
iris = pd.read_csv("iris.csv")
self.iris = Dataset(iris)
# Regression problem
tips = pd.read_csv("tips.csv")
self.tips = Dataset(tips, dependent_col="tip")
def test_number_of_rows(self):
result = self.iris.n_rows()
self.assertEqual(150, result)
def test_number_of_columns(self):
result = self.iris.n_columns()
self.assertEqual(5, result)
def test_if_self_categorical_cols_is_zero_iris(self):
result = self.iris.categorical_cols
self.assertEqual([], result)
def test_number_of_categorical_vars(self):
result = self.iris.n_categorical()
self.assertEqual(0, result)
def test_number_of_numerical_vars(self):
result = self.iris.n_numerical()
self.assertEqual(4, result)
def test_total_nclasses(self):
result = self.iris.n_classes()
self.assertEqual(3, result)
def test_total_nclasses_in_regression_problem(self):
result = self.tips.n_classes()
self.assertTrue(math.isnan(result))
def test_prediction_type_classification(self):
result = self.iris.prediction_type
self.assertEqual("classification", result)
def test_prediction_type_regression(self):
result = self.tips.prediction_type
self.assertEqual("regression", result)
def test_max_corr_with_dependent_classification(self):
result = self.iris.corr_with_dependent_abs_max()
self.assertTrue(math.isnan(result))
def test_max_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_max()
self.assertAlmostEqual(0.675, result, places=2)
def test_min_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_min()
self.assertAlmostEqual(0.002, result, places=2)
def test_25p_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_25p()
self.assertAlmostEqual(0.055, result, places=2)
def test_mean_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_mean()
self.assertAlmostEqual(0.18, result, places=2)
def test_median_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_median()
self.assertAlmostEqual(0.095, result, places=2)
def test_75p_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_75p()
self.assertAlmostEqual(0.125, result, places=2)
def test_std_corr_with_dependent_regression(self):
result = self.tips.corr_with_dependent_abs_std()
self.assertAlmostEqual(0.234, result, places=2)
def test_class_probablity_total_suite(self):
## evaluating for regression problem.
result = self.tips.class_prob_min()
self.assertTrue(math.isnan(result))
## evaluating for iris dataset.
## not really a useful test it seems, should get another dataset.
## Also not a unit test.
self.assertAlmostEqual(0.33, self.iris.class_prob_min(), places=2)
self.assertAlmostEqual(0.33, self.iris.class_prob_max(), places=2)
self.assertAlmostEqual(0, self.iris.class_prob_std(), places=2)
self.assertAlmostEqual(0.33, self.iris.class_prob_mean(), places=2)
self.assertAlmostEqual(0.33, self.iris.class_prob_median(), places=2)
def test_class_symbols_suite(self):
mean = 2.5
std = 1
min_ = 2
max_ = 4
## Iris has no nominal variables.
result = self.iris.symbols_sum()
self.assertTrue(math.isnan(result))
## evaluating for tips dataset.
## Also not a unit test.
symbol_counts = self.tips._get_symbols_per_category()
self.assertEqual(symbol_counts["sex"], 2)
self.assertEqual(symbol_counts["smoker"], 2)
self.assertAlmostEqual(max_, self.tips.symbols_max(), places=2)
self.assertAlmostEqual(min_, self.tips.symbols_min(), places=2)
self.assertAlmostEqual(std, self.tips.symbols_std(), places=2)
self.assertAlmostEqual(mean, self.tips.symbols_mean(), places=2)