def setUpClass(cls):
"""Setup the model only once, since it is expensive and can be copied"""
cls.data = crabs.mixed()
cls.model = MixableCondGaussianModel('crabs').fit(cls.data)
cls.model.mode = 'model'
# crabs has columns: 'species', 'sex', 'FL', 'RW', 'CL', 'CW', 'BD'
cls.cols = list(cls.data.columns)
class TestParallelProcessing(unittest.TestCase):
def setUp(self):
self.data = test_iris.mixed()
self.model = MixableCondGaussianModel("TestMod")
self.model.fit(df=self.data, fit_algo="map")
pass
def test_prob(self):
pred = ['sepal_width', 'sepal_length', Aggregation(['sepal_length', 'sepal_width'], method='probability', yields=None, args=None)]
# TODO: use Split instead of SplitTuple
split = [SplitTuple(name='sepal_width', method='equiinterval', args=[10]), SplitTuple(name='sepal_length', method='equiinterval', args=[10])]
self.model.parallel_processing = True
df_parallel = self.model.predict(predict=pred, splitby=split)
self.model.parallel_processing = False
df_serial = self.model.predict(predict=pred, splitby=split)
self.assertTrue(df_parallel.equals(df_serial))
def test_maximum(self):
# TODO: use Split instead of SplitTuple, and Aggregation instead of AggregationTuple
pred = ['sepal_width', 'sepal_length', AggregationTuple(name=['species'], method='maximum', yields='species', args=[])]
split = [SplitTuple(name='sepal_width', method='equiinterval', args=[10]), SplitTuple(name='sepal_length', method='equiinterval', args=[10])]
self.model.parallel_processing = True
df_parallel = self.model.predict(predict=pred, splitby=split)
self.model.parallel_processing = False
df_serial = self.model.predict(predict=pred, splitby=split)
self.assertTrue(df_parallel.equals(df_serial))
def test_samplequality(self):
samples1 = self.model._sample(250)
data1 = pd.DataFrame(data=samples1, columns=self.model.names)
testmod1 = MixCondGauss("Allbus_test1")
testmod1.fit(df=data1)
error1 = ((np.array(testmod1._mu).ravel() -
np.array(self.model._mu).ravel())**2).sum()
samples2 = self.model._sample(500)
data2 = pd.DataFrame(data=samples2, columns=self.model.names)
testmod2 = MixCondGauss("Allbus_test2")
testmod2.fit(df=data2)
error2 = ((np.array(testmod2._mu).ravel() -
np.array(self.model._mu).ravel())**2).sum()
self.assertTrue(error2 < error1)
class TestMethods(unittest.TestCase):
def setUp(self):
self.data = ta.mixed()
self.model = MixCondGauss("TestMod")
self.model.fit(df=self.data)
pass
def test_basics(self):
self.assertEqual(self.model.name, 'TestMod')
self.assertEqual(len(self.model._sample(5)), 5)
def test_samplequality(self):
samples1 = self.model._sample(250)
data1 = pd.DataFrame(data=samples1, columns=self.model.names)
testmod1 = MixCondGauss("Allbus_test1")
testmod1.fit(df=data1)
error1 = ((np.array(testmod1._mu).ravel() -
np.array(self.model._mu).ravel())**2).sum()
samples2 = self.model._sample(500)
data2 = pd.DataFrame(data=samples2, columns=self.model.names)
testmod2 = MixCondGauss("Allbus_test2")
testmod2.fit(df=data2)
error2 = ((np.array(testmod2._mu).ravel() -
np.array(self.model._mu).ravel())**2).sum()
self.assertTrue(error2 < error1)
def test_onlycats(self):
margmod = self.model.copy().marginalize(keep=['sex'])
self.assertEqual(type(margmod._sample(1)[0][0]), type('str'))
def test_onlynumericals(self):
margmod = self.model.copy().marginalize(keep=['age'])
self.assertEqual(type(margmod._sample(1)[0][0]), type(0.12345))
def setUp(self):
self.data = test_iris.mixed()
self.model = MixableCondGaussianModel("TestMod")
self.model.fit(df=self.data, fit_algo="map")
pass
def setUp(self):
self.data = ta.mixed()
self.model = MixCondGauss("TestMod")
self.model.fit(df=self.data)
pass
res = model.predict(['sex', 'RW', Aggregation(FL, method='maximum', yields='FL')], splitby=[Split(sex), Split(RW, method='equidist')])
print(res)
res = model.predict(['species', Density(species)], splitby=Split(species))
print(res)
res = model.predict(['FL', Density(FL)], splitby=Split(FL, method='equidist'))
print(res)
res = model.predict([Aggregation(FL, method='maximum', yields='FL')])
print(res)
res = model.predict([Aggregation(FL, method='maximum', yields='FL')], splitby=[Split(sex)])
print(res)
res = model.predict(['sex', Aggregation(FL, method='maximum', yields='FL')], splitby=[Split(sex)])
print(res)
res = model.predict(['FL', Probability(FL)], splitby=Split(FL, method='equiinterval'))
print(res)
res = model.predict(['sex', Density(sex)], splitby=Split(sex))
print(res)
res = model.predict(['sex', Density(sex), Aggregation(FL, method='maximum', yields='FL')], splitby=[Split(sex)])
print(res)
model = CondGaussian("my_model").fit(df=data)
res = model.predict(['sex', Density(sex)], splitby=Split(sex))
print(res)