def compare_with_default_hyper_params(prj):
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1, use_cache=False, hyper_params=prj['params'])
l.get_predicted_data()
l.print_scores()
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1, use_cache=False, hyper_params={})
l.get_predicted_data()
l.print_scores()
print()
def get_number(prj, num_of_learn_days):
l = IncrementalLearningModel(prj, 'DecisionTree', num_of_learn_days, 1)
y_proba, y_test = l.get_predicted_data()
b = RiskIsolate(y_proba, y_test, test_all_after_size_4=False)
b.get_num_of_exec_per_batch_size()
return b.get_max_improvement()
def batch_stop_4():
output = "\n{:28} | {} | {}\n".format('Project',
'Number of build execution',
'Improvement over TestAll')
output += '-' * 83
output += '\n'
print('Creating learning model for each project ...')
for prj in project_list:
l = IncrementalLearningModel(prj['name'],
'RF',
30,
1,
hyper_params={'n_estimators': 10})
y_proba, y_test = l.get_predicted_data()
b = BatchBisect(y_test, stop_at_4=True, batch_size_max=8)
b.get_num_of_exec_per_batch_size()
output += '{:28} | {} | {} \n'.format(
prj['name'],
str(b.lowest_num_of_exec).center(25),
'{:.2f} %'.format(b.max_improvement).center(22))
print(output)
def get_results(prj):
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1)
y_proba, y_test = l.get_predicted_data()
b = BatchBisect(y_test, stop_at_4=False, batch_size_max=20)
b.get_num_of_exec_per_batch_size()
return b
def get_results(prj):
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1)
y_proba, y_test = l.get_predicted_data()
b = RiskTopN(y_proba, y_test, top_n=2, batch_size_max=8)
b.get_num_of_exec_per_batch_size()
return b
def get_number(prj):
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1)
y_proba, y_test = l.get_predicted_data()
b = BatchBisect(y_test, no_bisect=True)
x, y = b.get_num_of_exec_per_batch_size()
print(b.max_improvement)
return x, y
def get_number(prj):
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1)
y_proba, y_test = l.get_predicted_data()
# l.print_scores()
b = RiskTopN(y_proba, y_test, top_n=2)
x, y = b.get_num_of_exec_per_batch_size()
print(b.lowest_num_of_exec)
return x, y
def find_optimum_hyper_params_for_random_forest(prj):
for n_estimators in [10, 50, 100, 200, 400]:
hyper_params = {'n_estimators': n_estimators}
for _ in ['RF']:
l = IncrementalLearningModel(prj, _, 30, 1, use_cache=False, hyper_params=hyper_params)
l.get_predicted_data()
l.print_scores()
if prj['params']['n_estimators'] == 10:
return
for max_depth in [10, 20, 50, 100, 200, None]:
hyper_params = {'max_depth': max_depth, 'n_estimators': prj['params']['n_estimators']}
for _ in ['RF']:
l = IncrementalLearningModel(prj['name'], _, 30, 1, use_cache=False, hyper_params=hyper_params)
l.get_predicted_data()
l.print_scores()
for criterion in ['gini', 'entropy']:
hyper_params = {'criterion': criterion, 'n_estimators': prj['params']['n_estimators']}
for _ in ['RF']:
l = IncrementalLearningModel(prj['name'], _, 30, 1, use_cache=False, hyper_params=hyper_params)
l.get_predicted_data()
l.print_scores()
for min_samples_split in [2, 5, 10, 20, 50, 100]:
hyper_params = {'min_samples_split': min_samples_split, 'n_estimators': prj['params']['n_estimators']}
for _ in ['RF']:
l = IncrementalLearningModel(prj['name'], _, 30, 1, use_cache=False, hyper_params=hyper_params)
l.get_predicted_data()
l.print_scores()
for min_samples_leaf in [1, 2, 5, 10, 20, 50, 100]:
hyper_params = {'min_samples_leaf': min_samples_leaf, 'n_estimators': prj['params']['n_estimators']}
for _ in ['RF']:
l = IncrementalLearningModel(prj['name'], _, 30, 1, use_cache=False, hyper_params=hyper_params)
l.get_predicted_data()
l.print_scores()
def get_number(prj):
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1)
y_proba, y_test = l.get_predicted_data()
b = BatchBisect(y_test, stop_at_4=True, batch_size_max=8)
x, y = b.get_num_of_exec_per_batch_size()
# TestAll Num of Executions
print(len(y_test))
# BatchBisect Num of Execution
print()
# print(b.lowest_num_of_exec / min(b.batch2_num_of_exec, b.batch4_num_of_exec, b.batch6_num_of_exec, b.batch8_num_of_exec, b.batch10_num_of_exec))
return x, y
def test_all():
output = "\n{:28} | {}\n".format('Project', 'Number of build execution')
output += '-' * 56
output += '\n'
print('Creating learning model for each project ...')
for prj in project_list:
l = IncrementalLearningModel(prj['name'],
'RF',
30,
1,
hyper_params={'n_estimators': 10})
y_proba, y_test = l.get_predicted_data()
output += '{:28} | {}\n'.format(prj['name'],
str(len(y_test)).center(25))
print(output)
def get_number(prj, approach):
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1)
y_proba, y_test = l.get_predicted_data()
if approach == 'BatchBisect':
b = BatchBisect(y_test, stop_at_4=False, batch_size_max=MAX_BATCH_SIZE)
elif approach == 'BatchStop4':
b = BatchBisect(y_test, stop_at_4=True, batch_size_max=MAX_BATCH_SIZE)
elif approach == 'RiskTopN':
b = RiskTopN(y_proba,
y_test,
top_n=RISK_TOP_N,
batch_size_max=MAX_BATCH_SIZE)
else:
raise NotImplementedError
x, y = b.get_num_of_exec_per_batch_size()
print(b.lowest_num_of_exec)
return x, y
def get_y_test_y_proba(prj):
learning_model = IncrementalLearningModel(prj['name'], 'RF', 30, 1)
y_proba, y_test = learning_model.get_predicted_data()
y_test = [True if item == 'passed' else False for item in y_test]
return y_test, y_proba
def get_testing_dataset_size(prj):
l = IncrementalLearningModel(prj['name'], 'RF', 30, 1)
y_proba, y_test = l.get_predicted_data()
return len(y_test)
def compare_classifiers(prj):
for _ in ['DT', 'RF', 'NB', 'MLP', 'LR', 'SGD']:
l = IncrementalLearningModel(prj['name'], _, 30, 1, use_cache=False, hyper_params={})
l.get_predicted_data()
l.print_scores()