def test_asym_gain_ratio_default(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf, features = idadf.columns, symmetry=False)
assert(isinstance(result, pandas.core.frame.DataFrame))
assert(len(result.columns) == len(idadf.columns))
assert(len(result.index) == len(idadf.columns))
result2 = gain_ratio(idadf, symmetry=False)
assert(all(result == result2))
result = result.fillna(0) # np.nan values are not equal when compared
assert(any(result != result.T)) # asymmetry
def test_asym_gain_ratio_default(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf, features=idadf.columns, symmetry=False)
assert (isinstance(result, pandas.core.frame.DataFrame))
assert (len(result.columns) == len(idadf.columns))
assert (len(result.index) == len(idadf.columns))
result2 = gain_ratio(idadf, symmetry=False)
assert (all(result == result2))
result = result.fillna(
0) # np.nan values are not equal when compared
assert (any(result != result.T)) # asymmetry
def test_sym_gain_ratio_one_target_one_feature(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf,
target=idadf.columns[0],
features=[idadf.columns[1]],
symmetry=True)
assert (isinstance(result, float))
result2 = gain_ratio(idadf,
target=idadf.columns[1],
features=[idadf.columns[0]],
symmetry=True)
assert (round(result, 3) == round(result2, 3)) # symmetry
def test_asym_gain_ratio_valueError(self, idadf):
if len(idadf.columns) > 0:
with pytest.raises(
ValueError
): # Cannot compute correlation coefficients of only one column (...), need at least 2
gain_ratio(idadf, features=idadf.columns[0], symmetry=False)
with pytest.raises(
ValueError
): # The correlation value of two same columns is always maximal
gain_ratio(idadf,
target=idadf.columns[0],
features=idadf.columns[0],
symmetry=False)
def test_asym_gain_ratio_one_target_one_feature(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf,
target=idadf.columns[0],
features=[idadf.columns[1]],
symmetry=False)
assert (isinstance(result, float))
def test_asym_gain_ratio_multiple_target(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf,
target=[idadf.columns[0], idadf.columns[1]],
symmetry=False)
assert (isinstance(result, pandas.core.frame.DataFrame))
assert (len(result.columns) == 2)
assert (len(result.index) == len(idadf.columns))
def test_asym_gain_ratio_one_target_one_feature(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf, target = idadf.columns[0], features=[idadf.columns[1]], symmetry=False)
assert(isinstance(result, float))
def test_asym_gain_ratio_multiple_target(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf, target = [idadf.columns[0],idadf.columns[1]], symmetry=False)
assert(isinstance(result, pandas.core.frame.DataFrame))
assert(len(result.columns) == 2)
assert(len(result.index) == len(idadf.columns))
def test_asym_gain_ratio_one_target(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf, target = idadf.columns[0], symmetry=False)
assert(isinstance(result, pandas.core.series.Series))
assert(len(result) == (len(idadf.columns)-1))
def test_asym_gain_ratio_valueError(self, idadf):
if len(idadf.columns) > 0:
with pytest.raises(ValueError): # Cannot compute correlation coefficients of only one column (...), need at least 2
gain_ratio(idadf, features = idadf.columns[0], symmetry=False)
with pytest.raises(ValueError): # The correlation value of two same columns is always maximal
gain_ratio(idadf, target= idadf.columns[0], features = idadf.columns[0], symmetry=False)
def test_sym_gain_ratio_one_target_one_feature(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf, target = idadf.columns[0], features=[idadf.columns[1]], symmetry=True)
assert(isinstance(result, float))
result2 = gain_ratio(idadf, target = idadf.columns[1], features=[idadf.columns[0]], symmetry=True)
assert(round(result,3) == round(result2,3)) # symmetry
def test_asym_gain_ratio_one_target(self, idadf):
if len(idadf.columns) > 1:
result = gain_ratio(idadf, target=idadf.columns[0], symmetry=False)
assert (isinstance(result, pandas.core.series.Series))
assert (len(result) == (len(idadf.columns) - 1))