def test_transform_default(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
transformed = pn.data.transform(features)
norms = np.linalg.norm(np.float64(transformed.values), axis=1)
for i in range(len(norms)):
self.assertAlmostEqual(norms[i], 1.0)
def test_transform_default(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
transformed = pn.data.transform(features)
norms = np.linalg.norm(np.float64(transformed.values), axis=1)
for i in range(len(norms)):
self.assertAlmostEqual(norms[i], 1.0)
def test_transform_rows_vector(self):
n_sessions = 3
norm = 3.14159
features = pn.featurize(self.equity_data, n_sessions)
transformed = pn.data.transform(features, method="vector", norm=norm)
norms = np.linalg.norm(np.float64(transformed.values), axis=1)
for i in range(len(norms)):
self.assertAlmostEqual(norms[i], norm)
def test_transform_rows_vector(self):
n_sessions = 3
norm = 3.14159
features = pn.featurize(self.equity_data, n_sessions)
transformed = pn.data.transform(features, method="vector", norm=norm)
norms = np.linalg.norm(np.float64(transformed.values), axis=1)
for i in range(len(norms)):
self.assertAlmostEqual(norms[i], norm)
def test_featurize_selection(self):
n_sessions = 4
features = pn.featurize(self.equity_data,
n_sessions,
selection='Volume')
self.assertEqual(len(features.index),
len(self.equity_data.index) - n_sessions + 1)
self.assertEqual(len(features.columns), n_sessions)
def test_transform_rows_mean(self):
n_sessions = 3
norm = 10.0
features = pn.featurize(self.equity_data, n_sessions)
transformed = pn.data.transform(features, method="mean", norm=norm, axis=0)
means = transformed.mean(axis=1)
for i in range(len(means)):
self.assertAlmostEqual(means.iloc[i], norm)
def test_featurize_columns(self):
cols = ['one', 'two', 'three']
n_sessions = len(cols)
features = pn.featurize(self.equity_data, n_sessions, columns=cols)
self.assertEqual(len(features.index), len(self.equity_data.index) - n_sessions + 1)
self.assertEqual(len(features.columns), n_sessions)
for i in range(len(cols)):
self.assertEqual(cols[i], features.columns.values[i])
def test_featurize_columns(self):
cols = ['one', 'two', 'three']
n_sessions = len(cols)
features = pn.featurize(self.equity_data, n_sessions, columns=cols)
self.assertEqual(len(features.index),
len(self.equity_data.index) - n_sessions + 1)
self.assertEqual(len(features.columns), n_sessions)
for i in range(len(cols)):
self.assertEqual(cols[i], features.columns.values[i])
def test_transform_rows_last(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
labels = pd.DataFrame(features.iloc[:, -1] * 0.5, index=features.index)
transformed_features, transformed_labels = pn.data.transform(features, method="last", labels=labels)
for i in range(len(transformed_features.index)):
self.assertAlmostEqual(transformed_features.iloc[i, -1], 1.0)
for i in range(len(transformed_labels.index)):
self.assertAlmostEqual(transformed_labels.iloc[i, 0], 0.5)
def test_transform_rows_last(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
labels = pd.DataFrame(features.iloc[:, -1] * 0.5, index=features.index)
transformed_features, transformed_labels = pn.data.transform(
features, method="last", labels=labels)
for i in range(len(transformed_features.index)):
self.assertAlmostEqual(transformed_features.iloc[i, -1], 1.0)
for i in range(len(transformed_labels.index)):
self.assertAlmostEqual(transformed_labels.iloc[i, 0], 0.5)
def test_transform_rows_mean(self):
n_sessions = 3
norm = 10.0
features = pn.featurize(self.equity_data, n_sessions)
transformed = pn.data.transform(features,
method="mean",
norm=norm,
axis=0)
means = transformed.mean(axis=1)
for i in range(len(means)):
self.assertAlmostEqual(means.iloc[i], norm)
def test_add_const_df(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
x = pn.data.feat.add_const(features)
self.assertTrue(isinstance(x, pd.DataFrame))
self.assertFalse(isinstance(x, np.ndarray))
self.assertEqual(len(x.index), len(features.index))
self.assertEqual(len(x.columns), len(features.columns) + 1)
for i in range(len(x.index)):
self.assertAlmostEqual(x.iloc[i, 0], 1.0)
for j in range(len(features.columns)):
self.assertAlmostEqual(x.iloc[i, j + 1], features.iloc[i, j])
def test_add_const_ndarray(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions).values
x = pn.data.feat.add_const(features)
self.assertTrue(isinstance(x, np.ndarray))
self.assertFalse(isinstance(x, pd.DataFrame))
self.assertEqual(x.shape[0], features.shape[0])
self.assertEqual(x.shape[1], features.shape[1] + 1)
for i in range(x.shape[0]):
self.assertAlmostEqual(x[i, 0], 1.0)
for j in range(features.shape[1]):
self.assertAlmostEqual(x[i, j + 1], features[i, j])
def test_transform_rows_first(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
# 2 columns of labels
labels = pd.DataFrame(index=features.index, columns=['Label1', 'Label2'])
labels['Label1'] = features.iloc[:, 0] * 0.5
labels['Label2'] = features.iloc[:, 0] * 2.0
transformed_features, transformed_labels = pn.data.transform(features, method="first", labels=labels)
for i in range(len(transformed_features.index)):
self.assertAlmostEqual(transformed_features.iloc[i, 0], 1.0)
for i in range(len(transformed_labels.index)):
self.assertAlmostEqual(transformed_labels.iloc[i, 0], 0.5)
self.assertAlmostEqual(transformed_labels.iloc[i, 1], 2.0)
def test_transform_rows_first(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
# 2 columns of labels
labels = pd.DataFrame(index=features.index,
columns=['Label1', 'Label2'])
labels['Label1'] = features.iloc[:, 0] * 0.5
labels['Label2'] = features.iloc[:, 0] * 2.0
transformed_features, transformed_labels = pn.data.transform(
features, method="first", labels=labels)
for i in range(len(transformed_features.index)):
self.assertAlmostEqual(transformed_features.iloc[i, 0], 1.0)
for i in range(len(transformed_labels.index)):
self.assertAlmostEqual(transformed_labels.iloc[i, 0], 0.5)
self.assertAlmostEqual(transformed_labels.iloc[i, 1], 2.0)
def test_featurize_defaults(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
self.assertEqual(len(features.index), len(self.equity_data.index) - n_sessions + 1)
self.assertEqual(len(features.columns), n_sessions)
def test_featurize_defaults(self):
n_sessions = 3
features = pn.featurize(self.equity_data, n_sessions)
self.assertEqual(len(features.index),
len(self.equity_data.index) - n_sessions + 1)
self.assertEqual(len(features.columns), n_sessions)
def test_featurize_selection(self):
n_sessions = 4
features = pn.featurize(self.equity_data, n_sessions, selection='Volume')
self.assertEqual(len(features.index), len(self.equity_data.index) - n_sessions + 1)
self.assertEqual(len(features.columns), n_sessions)
