def test_csv_merge(self):
X_train, y_train, X_test, y_test = get_dev_split(raw)
X_test_ext = get_extended_X(X_test, raw)
# Check join with tube.csv.
taid = 'TA-00034'
df = X_test_ext[X_test['tube_assembly_id'] == taid]
ref_diams = raw['tube'][raw['tube']['tube_assembly_id'] == taid]
nose.tools.assert_equal(len(df['diameter'].unique()), 1)
nose.tools.assert_equal(len(ref_diams), 1)
nose.tools.assert_equal(
df['diameter'].unique()[0],
ref_diams['diameter'].values[0])
# Check join with specs.csv.
taid = 'TA-00207'
ref_specs = ['SP-0063', 'SP-0070', 'SP-0080']
df = X_test_ext[X_test['tube_assembly_id'] == taid]
assert len(df['specs']) >= 1
for val in df['specs']:
nose.tools.assert_equal(val, ref_specs)
# Check join with bill_of_materials.csv.
taid = 'TA-00249'
ref_components = [
('C-1536', 2.0),
('C-1642', 1.0),
('C-1649', 1.0),
]
df = X_test_ext[X_test['tube_assembly_id'] == taid]
assert len(df['components']) >= 1
for val in df['components']:
nose.tools.assert_equal(val, ref_components)
def test_csv_merge(self):
X_train, y_train, X_test, y_test = get_dev_split(raw)
X_test_ext = get_extended_X(X_test, raw)
# Check join with tube.csv.
taid = 'TA-00034'
df = X_test_ext[X_test['tube_assembly_id'] == taid]
ref_diams = raw['tube'][raw['tube']['tube_assembly_id'] == taid]
nose.tools.assert_equal(len(df['diameter'].unique()), 1)
nose.tools.assert_equal(len(ref_diams), 1)
nose.tools.assert_equal(df['diameter'].unique()[0],
ref_diams['diameter'].values[0])
# Check join with specs.csv.
taid = 'TA-00207'
ref_specs = ['SP-0063', 'SP-0070', 'SP-0080']
df = X_test_ext[X_test['tube_assembly_id'] == taid]
assert len(df['specs']) >= 1
for val in df['specs']:
nose.tools.assert_equal(val, ref_specs)
# Check join with bill_of_materials.csv.
taid = 'TA-00249'
ref_components = [
('C-1536', 2.0),
('C-1642', 1.0),
('C-1649', 1.0),
]
df = X_test_ext[X_test['tube_assembly_id'] == taid]
assert len(df['components']) >= 1
for val in df['components']:
nose.tools.assert_equal(val, ref_components)
def test_custom_features(self):
X_train, y_train, X_test, y_test = get_dev_split(raw)
X_train_ext = get_extended_X(X_train, raw)
featurizer = CustomFeaturizer()
featurizer.fit(X_train_ext)
X_train_feats = featurizer.transform(X_train_ext)
# Check adj_quantity feature.
taid = 'TA-01916'
df = X_train_feats[X_train['tube_assembly_id'] == taid]
nose.tools.assert_equal(len(df), 2)
nose.tools.assert_true(np.all(df.min_order_quantity == [1, 4]))
nose.tools.assert_true(np.all(df.quantity == [1, 1]))
nose.tools.assert_true(np.all(df.adj_quantity == [1, 4]))
def test_custom_features(self):
X_train, y_train, X_test, y_test = get_dev_split(raw)
X_train_ext = get_extended_X(X_train, raw)
featurizer = CustomFeaturizer()
featurizer.fit(X_train_ext)
X_train_feats = featurizer.transform(X_train_ext)
# Check adj_quantity feature.
taid = 'TA-01916'
df = X_train_feats[X_train['tube_assembly_id'] == taid]
nose.tools.assert_equal(len(df), 2)
nose.tools.assert_true(np.all(df.min_order_quantity == [1, 4]))
nose.tools.assert_true(np.all(df.quantity == [1, 1]))
nose.tools.assert_true(np.all(df.adj_quantity == [1, 4]))
def test_dev_split_no_overlap(self):
X_train, y_train, X_test, y_test = get_dev_split(raw)
# All instances assigned to either test or train.
nose.tools.assert_equal(len(X_train), len(y_train))
nose.tools.assert_equal(len(X_test), len(y_test))
nose.tools.assert_equal(
len(X_train) + len(X_test), len(raw['train_set']))
# Test and train contain no `tube_assembly_id`s in common.
train_ids = set(X_train['tube_assembly_id'])
test_ids = set(X_test['tube_assembly_id'])
nose.tools.assert_false(train_ids.intersection(test_ids))
# Split is roughly 90% / 10%.
test_frac = 1.0 * len(X_test) / len(raw['train_set'])
print test_frac
nose.tools.assert_almost_equals(test_frac, 0.1, delta=0.05)
def test_dev_split_no_overlap(self):
X_train, y_train, X_test, y_test = get_dev_split(raw)
# All instances assigned to either test or train.
nose.tools.assert_equal(len(X_train), len(y_train))
nose.tools.assert_equal(len(X_test), len(y_test))
nose.tools.assert_equal(
len(X_train) + len(X_test), len(raw['train_set']))
# Test and train contain no `tube_assembly_id`s in common.
train_ids = set(X_train['tube_assembly_id'])
test_ids = set(X_test['tube_assembly_id'])
nose.tools.assert_false(train_ids.intersection(test_ids))
# Split is roughly 90% / 10%.
test_frac = 1.0 * len(X_test) / len(raw['train_set'])
print test_frac
nose.tools.assert_almost_equals(test_frac, 0.1, delta=0.05)
