def test_summary():
clf = LogisticRegression()
data = ModelingData.from_dataframe(df, target='group')
data.fit(clf)
probas = data.predict_proba(clf)
summary = ModelingData.get_threshold_summary(probas, 1)
assert_dict_equal(
dict(summary), {
'sensiticity': 1.0,
'target': 1,
'f1': 1.0,
'recall': 1.0,
'num_negatives': 7,
'num_positives': 5,
'false_positive_rate': 0.0,
'false_positives': 0,
'precision': 1.0,
'true_positives': 5,
'false_negatives': 0,
'true_positive_rate': 1.0,
'specificity': 1.0,
'threshold': 0.5,
'true_negatives': 7,
'accuracy': 1.0
})
def test_orthogonal():
dataA = ModelingData.from_dataframe(df.iloc[0:6], target='group')
dataB = ModelingData.from_dataframe(df.iloc[3:9], target='group')
dataC = ModelingData.from_dataframe(df.iloc[6:12], target='group')
assert (dataA.is_orthogonal(dataC))
assert (not dataA.is_orthogonal(dataB))
assert (not dataB.is_orthogonal(dataC))
def test_classifier_performance_summary():
clf = LogisticRegression()
data = ModelingData.from_dataframe(df, target='group')
data.fit(clf)
probas, summary = data.get_classifier_performance_summary(
clf, 0, thresholds=np.arange(0.0, 1.0, 0.1))
assert_dict_equal(
dict(summary.irow(0)), {
'f1': 0.7368421052631579,
'target': 0.0,
'sensiticity': 1.0,
'recall': 1.0,
'num_negatives': 0.0,
'num_positives': 12.0,
'false_positive_rate': 1.0,
'false_positives': 5.0,
'precision': 0.58333333333333337,
'true_positives': 7.0,
'false_negatives': 0.0,
'true_positive_rate': 1.0,
'specificity': 0.0,
'true_negatives': 0.0,
'accuracy': 0.58333333333333337
})
def test_numeric_features():
data = ModelingData.from_dataframe(df2, target='group')
eq_(data.shape(), (12, 3))
assert ('feature3' in data.features.columns)
numeric_data = data.numeric_features()
eq_(numeric_data.shape(), (12, 2))
assert ('feature3' not in numeric_data.features.columns)
def test_split():
random_state = RandomState(12345)
data = ModelingData.from_dataframe(df, target='group')
train, test = data.train_test_split(random_state=random_state)
print train, test
eq_(train.shape(), (9, 2))
eq_(test.shape(), (3, 2))
assert (train.is_orthogonal(test))
def test_probas():
clf = LogisticRegression()
data = ModelingData.from_dataframe(df, target='group')
data.fit(clf)
probas = data.predict_proba(clf)
eq_(
dict(probas.irow(0)), {
'index': 0.0,
'proba_0': 0.58298495576378184,
'proba_1': 0.41701504423621821,
'target': 0.0
})
def test_balance_truncation():
data = ModelingData.from_dataframe(df, target='group')
eq_(len(data), 12)
value_counts = data.targets.value_counts()
eq_(value_counts[0], 7)
eq_(value_counts[1], 5)
# Now, balance the data
balanced = data._balance_by_truncation()
eq_(len(balanced), 10)
value_counts = balanced.targets.value_counts()
eq_(value_counts[0], 5)
eq_(value_counts[1], 5)
def test_balance_sample():
data = ModelingData.from_dataframe(df, target='group')
eq_(len(data), 12)
value_counts = data.targets.value_counts()
eq_(value_counts[0], 7)
eq_(value_counts[1], 5)
# Now, balance the data
np.random.seed(42)
balanced = data._balance_by_sample_with_replace(size=20)
eq_(len(balanced), 20)
value_counts = balanced.targets.value_counts()
eq_(value_counts[0], 10)
eq_(value_counts[1], 10)
def test_predict():
reg = LinearRegression()
data = ModelingData.from_dataframe(df, target='group')
data.fit(reg)
predictions = data.predict(reg)
eq_(dict(predictions.irow(0)), {
'predict': 0.15360473177939843,
'index': 0.0,
'target': 0.0
})
eq_(dict(predictions.irow(-1)), {
'predict': 1.2116420074937881,
'index': 11.0,
'target': 1.0
})