def test_error_on_mismatching_classifiers(self):
"""check that an error is raised on mismatch in the number of
classifiers
"""
classifiers_config = get_mock_classifiers_config()
rule_filter = ClassifierFilter(classifiers_config, [])
with self.assertRaises(ValueError):
rule_filter.create_classifiers_filter(
["c1_v1", "?", "agg1", "extra"], pd.DataFrame([[1, 3, 5]]))
with self.assertRaises(ValueError):
rule_filter.create_classifiers_filter(
["c1_v1", "?"], # one too few
pd.DataFrame([[1, 3, 5]]))
with self.assertRaises(ValueError):
rule_filter.create_classifiers_filter(
["c1_v1", "?", "agg1"],
pd.DataFrame([[1, 3]])) # not enough columns in dataframe
with self.assertRaises(ValueError):
rule_filter.create_classifiers_filter(
["c1_v1", "?", "agg1"],
pd.DataFrame([[1, 3, 5, 1]])) # too many columns in dataframe
def test_error_on_undefined_value_in_classifier_set(self):
"""check that an error is raised on when classifier value in
the specified classifier set is not defined
"""
classifiers_config = get_mock_classifiers_config()
rule_filter = ClassifierFilter(classifiers_config, [])
with self.assertRaises(ValueError):
rule_filter.create_classifiers_filter(["undefined", "?", "agg1"],
pd.DataFrame([[1, 3, 5]]))
def sit_rule_based_processor_factory(cbm, random_func, classifiers_config,
classifier_aggregates, sit_events,
sit_transitions, tr_constants,
sit_disturbance_eligibilities,
reset_parameters):
classifier_filter = ClassifierFilter(
classifiers_config=classifiers_config,
classifier_aggregates=classifier_aggregates)
state_filter_func = \
sit_transition_rule_processor.state_variable_filter_func
tr_processor = sit_transition_rule_processor.SITTransitionRuleProcessor(
TransitionRuleProcessor(
classifier_filter_builder=classifier_filter,
state_variable_filter_func=state_filter_func,
classifiers_config=classifiers_config,
grouped_percent_err_max=tr_constants.group_err_max,
wildcard=tr_constants.wildcard,
transition_classifier_postfix=tr_constants.classifier_value_postfix
))
event_processor = sit_event_processor.SITEventProcessor(
cbm=cbm,
classifier_filter_builder=classifier_filter,
random_generator=random_func)
return SITRuleBasedProcessor(event_processor, tr_processor, sit_events,
sit_transitions,
sit_disturbance_eligibilities,
reset_parameters)
def test_create_classifiers_filter_expected_value(self):
classifier_set = ["c1_v1", "?", "agg1"]
classifiers_config = get_mock_classifiers_config()
classifier_aggregates = [{
'classifier_id': 3,
'name': 'agg1',
'description': 'agg1',
'classifier_values': ['c3_v1', 'c3_v3']
}]
rule_filter = ClassifierFilter(classifiers_config,
classifier_aggregates)
def get_classifier_value_index(classifier_id):
return {
x["value"]: x["id"]
for x in classifiers_config["classifier_values"]
if x["classifier_id"] == classifier_id
}
c1 = get_classifier_value_index(1)
c2 = get_classifier_value_index(2)
c3 = get_classifier_value_index(3)
classifier_values = pd.DataFrame(
[
(c1[x[0]], c2[x[1]], c3[x[2]]) for x in [
("c1_v1", "c2_v1", "c3_v3"), # match
("c1_v2", "c2_v1", "c3_v3"), # non-match (c1_v2)
("c1_v1", "c2_v2", "c3_v3"), # match
("c1_v1", "c2_v2", "c3_v1"), # match
("c1_v1", "c2_v2", "c3_v2")
] # non-match (aggregate)
],
columns=["c1", "c2", "c3"])
result = rule_filter.create_classifiers_filter(classifier_set,
classifier_values)
self.assertTrue(
result.expression == "(c_0 == 1) & ((c_2 == 5) | (c_2 == 7))")
self.assertTrue(list(result.local_dict["c_0"]) == [1, 2, 1, 1, 1])
self.assertTrue(list(result.local_dict["c_1"]) == [3, 3, 4, 4, 4])
self.assertTrue(list(result.local_dict["c_2"]) == [7, 7, 7, 5, 6])
def test_create_classifiers_filter_expected_value_all_wildcards(self):
classifier_set = ["?", "?", "?"]
classifiers_config = get_mock_classifiers_config()
classifier_aggregates = [{
'classifier_id': 3,
'name': 'agg1',
'description': 'agg1',
'classifier_values': ['c3_v1', 'c3_v3']
}]
rule_filter = ClassifierFilter(classifiers_config,
classifier_aggregates)
def get_classifier_value_index(classifier_id):
return {
x["value"]: x["id"]
for x in classifiers_config["classifier_values"]
if x["classifier_id"] == classifier_id
}
c1 = get_classifier_value_index(1)
c2 = get_classifier_value_index(2)
c3 = get_classifier_value_index(3)
classifier_values = pd.DataFrame(
[
(c1[x[0]], c2[x[1]], c3[x[2]]) for x in [
("c1_v1", "c2_v1", "c3_v3"), # match
("c1_v2", "c2_v1", "c3_v3"), # non-match (c1_v2)
("c1_v1", "c2_v2", "c3_v3"), # match
("c1_v1", "c2_v2", "c3_v1"), # match
("c1_v1", "c2_v2", "c3_v2")
] # non-match (aggregate)
],
columns=["c1", "c2", "c3"])
result = rule_filter.create_classifiers_filter(classifier_set,
classifier_values)
self.assertTrue(result.expression == "")
self.assertTrue(result.local_dict == {})
def get_transition_rules_pre_dynamics_func(sit):
classifiers_config = sit_cbm_factory.get_classifiers(
sit.sit_data.classifiers, sit.sit_data.classifier_values)
sit_transition_rules = sit_cbm_factory.initialize_transition_rules(sit)
classifier_filter = ClassifierFilter(
classifiers_config=classifiers_config,
classifier_aggregates=sit.sit_data.classifier_aggregates)
classifier_value_post_fix = \
sit_format.get_tr_classifier_set_postfix()
group_err_max = sit_transition_rule_parser.GROUPED_PERCENT_ERR_MAX
state_filter_func = \
sit_transition_rule_processor.state_variable_filter_func
tr_processor = TransitionRuleProcessor(
classifier_filter_builder=classifier_filter,
state_variable_filter_func=state_filter_func,
classifiers_config=classifiers_config,
grouped_percent_err_max=group_err_max,
wildcard=sit_classifier_parser.get_wildcard_keyword(),
transition_classifier_postfix=classifier_value_post_fix)
sit_tr_processor = \
sit_transition_rule_processor.SITTransitionRuleProcessor(tr_processor)
return sit_transition_rule_processor.get_pre_dynamics_func(
sit_tr_processor, sit_transition_rules)
def test_aggregate_value_interval_feature(self):
classifier_set = ["c1_v1", "?", "agg1"]
classifiers_config = {
"classifiers": [{
"id": 1,
"name": "c1"
}, {
"id": 2,
"name": "c2"
}, {
"id": 3,
"name": "c3"
}],
"classifier_values": [{
"id": 1,
"classifier_id": 1,
"value": "c1_v1"
}, {
"id": 2,
"classifier_id": 1,
"value": "c1_v2"
}, {
"id": 3,
"classifier_id": 2,
"value": "c2_v1"
}, {
"id": 4,
"classifier_id": 2,
"value": "c2_v2"
}, {
"id": 5,
"classifier_id": 3,
"value": "c3_v1"
}, {
"id": 6,
"classifier_id": 3,
"value": "c3_v2"
}, {
"id": 7,
"classifier_id": 3,
"value": "c3_v3"
}, {
"id": 8,
"classifier_id": 3,
"value": "c3_v4"
}, {
"id": 9,
"classifier_id": 3,
"value": "c3_v5"
}, {
"id": 10,
"classifier_id": 3,
"value": "c3_v6"
}]
}
classifier_aggregates = [{
'classifier_id':
3,
'name':
'agg1',
'description':
'agg1',
'classifier_values': ['c3_v1', 'c3_v2', 'c3_v4', 'c3_v6', 'c3_v3']
}]
# 5 6 8 10 7
classifier_values = pd.DataFrame(columns=["c1", "c2", "c3"])
rule_filter = ClassifierFilter(classifiers_config,
classifier_aggregates)
result = rule_filter.create_classifiers_filter(classifier_set,
classifier_values)
self.assertTrue(
result.expression ==
"(c_0 == 1) & (((c_2 >= 5) & (c_2 <= 8)) | (c_2 == 10))")