def given_a_sleeping_dog_test():
customer_segment = ('60602', 'male')
classification_model = SimplisticClasses.DumbClassifier({
('control', ) + customer_segment:
0.50,
('variant', ) + customer_segment:
0.45,
})
regression_model = SimplisticClasses.AllCasesHaveSameProfitRegressionModel(
)
ad_name = SimplisticClasses.assign_ad_for(customer_segment,
classification_model,
regression_model)
assert ad_name == 'control', "Should let sleeping dogs lie."
def given_a_variant_that_does_NOT_improve_on_probability_of_ordering_over_control_test(
):
customer_segment = ('60626', 'male')
classification_model = SimplisticClasses.DumbClassifier({
('control', ) + customer_segment:
0.45,
('variant', ) + customer_segment:
0.45,
})
regression_model = SimplisticClasses.AllCasesHaveSameProfitRegressionModel(
)
ad_name = SimplisticClasses.assign_ad_for(customer_segment,
classification_model,
regression_model)
assert ad_name == 'control', "Should choose to NOT advertise"
def given_variant_improves_over_control_just_enough_to_warrant_advertising_cost_test(
):
customer_segment = ('60626', 'female')
classification_model = SimplisticClasses.DumbClassifier({
('control', ) + customer_segment:
0.60,
('variant', ) + customer_segment:
0.65,
})
regression_model = SimplisticClasses.AllCasesHaveSameProfitRegressionModel(
)
ad_name = SimplisticClasses.assign_ad_for(customer_segment,
classification_model,
regression_model,
ad_cost=0.60)
assert ad_name == 'variant', "Should choose to advertise"
def given_a_dumb_classifer_that_says_what_I_want_test():
classifier = SimplisticClasses.DumbClassifier({
('control', '60626', 'female'):
0.60,
})
order_probability = classifier.probability(('control', '60626', 'female'))
assert order_probability == 0.60, "Should return probability I told it to."
def given_probability_to_order_remains_constant_but_expected_profit_increases_test(
):
customer_segment = ('60626', 'female')
classification_model = SimplisticClasses.DumbClassifier({
('control', ) + customer_segment:
0.65,
('variant', ) + customer_segment:
0.65,
})
regression_model = SimplisticClasses.DumbClassifier({
('control', ) + customer_segment:
12.25,
('variant', ) + customer_segment:
15.50,
})
ad_name = SimplisticClasses.assign_ad_for(customer_segment,
classification_model,
regression_model)
assert ad_name == 'variant', "Should recommend using ad"
def given_any_input_test():
regression_model = SimplisticClasses.AllCasesHaveSameProfitRegressionModel(
)
results = regression_model.predict(input=(42, 'hai'))
assert results == 12.25, "Should be a constant amount regardless of the input."
def linear_regression_test():
dummy_regression_model = DummyRegressionModel(value_predicted=33.12)
model = SimplisticClasses.RegressionModel(dummy_regression_model)
expected_profit_if_orders = model.predict([1, 5, 2])
assert expected_profit_if_orders == 33.12
def logistic_regression_test():
dummy_sklearn_model = DummySklearnModel(probability_of_ordering=0.42)
model = SimplisticClasses.LogisticModel(dummy_sklearn_model)
probability_of_ordering = model.probability([1, 2, 3])
assert dummy_sklearn_model.predict_proba_call_arguments == [1, 2, 3]
assert probability_of_ordering == 0.42
def given_a_never_before_seen_observation_test():
classifier = SimplisticClasses.DumbClassifier({})
probability = classifier.probability(('boo', 'bibbit'))
assert probability == None, "Should return None"