def select_optimal_features_set_using_univariate_feature_selection(
self, classes):
print "SELECTING OPTIMAL FEATURES SET USING RECURSIVE FEATURE ELIMINATION"
print reduce(
lambda result, class_name: update_and_return_json(
result, class_name,
adjust_optimal_features_using_recursive_feature_elimination(
class_name, self.mongoCollection.get_all_records())),
classes, {})
def select_optimal_features_set_using_univariate_feature_selection(self, classes):
print "SELECTING OPTIMAL FEATURES SET USING RECURSIVE FEATURE ELIMINATION"
print reduce(lambda result, class_name:
update_and_return_json(result,
class_name,
adjust_optimal_features_using_recursive_feature_elimination(class_name,
self.mongoCollection.get_all_records())),
classes,
{})
def benchmark(self, classes, classification_method, fields):
all_records = self.mongoCollection.get_all_records()
if all_records < 2:
return
print reduce(
lambda res, class_name: update_and_return_json(
res, class_name,
self.create_accuracy_ranking(all_records, fields, class_name,
classification_method)), classes,
{})
def benchmark(self, classes, classification_method, fields):
all_records = self.mongoCollection.get_all_records()
if all_records < 2:
return
print reduce(
lambda res, class_name: update_and_return_json(
res, class_name, self.create_accuracy_ranking(all_records, fields, class_name, classification_method)
),
classes,
{},
)
def create_accuracy_ranking(self, records, features_to_benchmark,
class_name, delivered_classification_method):
def increment_if_successful_classified(score, record, feature):
if record['classes'][
class_name] == delivered_classification_method(
record, class_name, feature,
[x for x in records if x != record]):
score += 1
return score
return reduce(
lambda result, feature: update_and_return_json(
result, feature,
reduce(
lambda score, record: increment_if_successful_classified(
score, record, feature), records, 0) /
(len(records) - 1)), features_to_benchmark, {})
def create_accuracy_ranking(self, records, features_to_benchmark, class_name, delivered_classification_method):
def increment_if_successful_classified(score, record, feature):
if record["classes"][class_name] == delivered_classification_method(
record, class_name, feature, [x for x in records if x != record]
):
score += 1
return score
return reduce(
lambda result, feature: update_and_return_json(
result,
feature,
reduce(lambda score, record: increment_if_successful_classified(score, record, feature), records, 0)
/ (len(records) - 1),
),
features_to_benchmark,
{},
)
