def bellwether(self):
final_score = {}
count = 0
for s_project in self.selected_projects:
try:
s_path = self.data_source + '/' + s_project
print(s_project)
df = self.prepare_data(s_path)
if df.shape[0] < 50:
continue
else:
count += 1
df.reset_index(drop=True, inplace=True)
d = {'buggy': True, 'clean': False}
df['Buggy'] = df['Buggy'].map(d)
y = df.Buggy
X = df.drop(labels=['Buggy'], axis=1)
kf = StratifiedKFold(n_splits=5)
score = {}
F = {}
for i in range(5):
for train_index, tune_index in kf.split(X, y):
X_train, X_tune = X.iloc[train_index], X.iloc[
tune_index]
y_train, y_tune = y[train_index], y[tune_index]
clf = LogisticRegression()
clf.fit(X_train, y_train)
destination_projects = copy.deepcopy(
self.selected_projects)
destination_projects.remove(s_project)
for d_project in destination_projects:
#print(d_project)
try:
d_path = self.data_source + '/' + d_project
dest_df = self.prepare_data(d_path)
if dest_df.shape[0] < 50:
continue
dest_df.reset_index(drop=True, inplace=True)
d = {'buggy': True, 'clean': False}
dest_df['Buggy'] = dest_df['Buggy'].map(d)
test_y = dest_df.Buggy
test_X = dest_df.drop(labels=['Buggy'], axis=1)
predicted = clf.predict(test_X)
_df_test_loc = test_X.LOC
abcd = metrices.measures(
test_y, predicted, _df_test_loc)
F['f1'] = [abcd.calculate_f1_score()]
F['precision'] = [abcd.calculate_precision()]
F['recall'] = [abcd.calculate_recall()]
F['g-score'] = [abcd.get_g_score()]
F['d2h'] = [abcd.calculate_d2h()]
F['pci_20'] = [abcd.get_pci_20()]
F['ifa'] = [abcd.get_ifa()]
F['pd'] = [abcd.get_pd()]
F['pf'] = [abcd.get_pf()]
_F = copy.deepcopy(F)
if 'f1' not in score.keys():
score[d_project] = _F
else:
score[d_project]['f1'].append(F['f1'][0])
score[d_project]['precision'].append(
F['precision'][0])
score[d_project]['recall'].append(
F['recall'][0])
score[d_project]['g-score'].append(
F['g-score'][0])
score[d_project]['d2h'].append(F['d2h'][0])
score[d_project]['pci_20'].append(
F['pci_20'][0])
score[d_project]['ifa'].append(F['ifa'][0])
score[d_project]['pd'].append(F['pd'][0])
score[d_project]['pf'].append(F['pf'][0])
except Exception as e:
print(e)
continue
final_score[s_project] = score
except Exception as e:
print(e)
continue
return final_score
def bellwether(self,selected_projects,all_projects):
final_score = {}
count = 0
for s_project in selected_projects:
try:
print(s_project,selected_projects.shape[0])
s_path = self.data_path + s_project
#print(s_project)
df = self.prepare_data(s_path)
if df.shape[0] < 50:
continue
else:
count+=1
df.reset_index(drop=True,inplace=True)
d = {'buggy': True, 'clean': False}
df['Buggy'] = df['Buggy'].map(d)
df, s_cols = self.apply_cfs(df)
#s_cols = df.columns.tolist()
df = self.apply_smote(df)
y = df.Buggy
X = df.drop(labels = ['Buggy'],axis = 1)
kf = StratifiedKFold(n_splits = 5)
score = {}
F = {}
for i in range(1):
clf = LogisticRegression()
# clf = RandomForestClassifier()
clf.fit(X,y)
destination_projects = copy.deepcopy(all_projects)
#destination_projects.remove(s_project)
for d_project in destination_projects:
try:
d_path = self.data_path + d_project
_test_df = self.prepare_data(d_path)
_df_test_loc = _test_df.LOC
test_df = _test_df[s_cols]
if test_df.shape[0] < 50:
continue
test_df.reset_index(drop=True,inplace=True)
d = {'buggy': True, 'clean': False}
test_df['Buggy'] = test_df['Buggy'].map(d)
test_y = test_df.Buggy
test_X = test_df.drop(labels = ['Buggy'],axis = 1)
predicted = clf.predict(test_X)
abcd = metrices.measures(test_y,predicted,_df_test_loc)
F['f1'] = [abcd.calculate_f1_score()]
F['precision'] = [abcd.calculate_precision()]
F['recall'] = [abcd.calculate_recall()]
F['g-score'] = [abcd.get_g_score()]
F['d2h'] = [abcd.calculate_d2h()]
F['pci_20'] = [abcd.get_pci_20()]
F['ifa'] = [abcd.get_ifa()]
F['pd'] = [abcd.get_pd()]
F['pf'] = [abcd.get_pf()]
_F = copy.deepcopy(F)
if 'f1' not in score.keys():
score[d_project] = _F
else:
score[d_project]['f1'].append(F['f1'][0])
score[d_project]['precision'].append(F['precision'][0])
score[d_project]['recall'].append(F['recall'][0])
score[d_project]['g-score'].append(F['g-score'][0])
score[d_project]['d2h'].append(F['d2h'][0])
score[d_project]['pci_20'].append(F['pci_20'][0])
score[d_project]['ifa'].append(F['ifa'][0])
score[d_project]['pd'].append(F['pd'][0])
score[d_project]['pf'].append(F['pf'][0])
except Exception as e:
print("dest",d_project,e)
continue
final_score[s_project] = score
except Exception as e:
print("src",s_project,e)
continue
return final_score
def get_predicted(cluster_data_loc, metrices_loc, fold, data_location,
default_bellwether_loc):
train_data = pd.read_pickle(cluster_data_loc + '/train_data.pkl')
cluster, cluster_tree, max_depth = cluster_driver(train_data)
t_df = pd.DataFrame()
for project in train_data.index.values.tolist():
_s_path = '/Users/suvodeepmajumder/Documents/AI4SE/bellwether_comminity/data/1385/converted/' + project
s_df = prepare_data(_s_path)
t_df = pd.concat([t_df, s_df])
break
t_df.reset_index(drop=True, inplace=True)
d = {'buggy': True, 'clean': False}
t_df['Buggy'] = t_df['Buggy'].map(d)
t_df, g_s_cols = apply_cfs(t_df)
t_df = apply_smote(t_df)
train_y = t_df.Buggy
train_X = t_df.drop(labels=['Buggy'], axis=1)
clf_global = LogisticRegression()
clf_global.fit(train_X, train_y)
test_data = pd.read_pickle(cluster_data_loc + '/test_data.pkl')
#print(test_data)
test_projects = test_data.index.values.tolist()
goals = ['recall', 'precision', 'pf', 'pci_20', 'ifa']
levels = [2, 1, 0]
results = {}
bellwether_models = {}
bellwether0_models = {}
bellwether0_s_cols = {}
bellwether_s_cols = {}
self_model = {}
self_model_test = {}
for level in levels:
test_data = test_data
predicted_cluster = cluster.predict(test_data, level)
#print(level,predicted_cluster)
for i in range(len(predicted_cluster)):
try:
F = {}
_F = {}
b_F = {}
g_F = {}
r_F = {}
c_id = predicted_cluster[i]
s_project_df = pd.read_csv(cluster_data_loc +
'/bellwether_cdom_' + str(level) +
'.csv')
if level == 1:
s_project_df.rename(columns={'Unnamed: 0': 'id'},
inplace=True)
if level == 0:
s_project = s_project_df.bellwether.values[0]
else:
s_project = s_project_df[
s_project_df['id'] ==
predicted_cluster[i]].bellwether.values[0]
if s_project not in bellwether_models.keys():
s_path = '/Users/suvodeepmajumder/Documents/AI4SE/bellwether_comminity/data/1385/converted/' + s_project
df = prepare_data(s_path)
df.reset_index(drop=True, inplace=True)
d = {'buggy': True, 'clean': False}
df['Buggy'] = df['Buggy'].map(d)
df, s_cols = apply_cfs(df)
bellwether_s_cols[s_project] = s_cols
df = apply_smote(df)
y = df.Buggy
X = df.drop(labels=['Buggy'], axis=1)
clf_bellwether = LogisticRegression()
clf_bellwether.fit(X, y)
bellwether_models[s_project] = clf_bellwether
else:
clf_bellwether = bellwether_models[s_project]
s_cols = bellwether_s_cols[s_project]
b_s_project_df = pd.read_csv(default_bellwether_loc +
'/cdom_bellwether.csv')
b_s_project_df.columns = [
'bellwether', 'recall', 'precision', 'pf', 'wins'
]
b_s_project = b_s_project_df.bellwether.values[
b_s_project_df.wins.idxmax()]
if b_s_project not in bellwether0_models.keys():
b_s_path = '/Users/suvodeepmajumder/Documents/AI4SE/bellwether_comminity/data/1385/converted/' + b_s_project
b_df = prepare_data(b_s_path)
b_df.reset_index(drop=True, inplace=True)
d = {'buggy': True, 'clean': False}
b_df['Buggy'] = b_df['Buggy'].map(d)
b_df, b_s_cols = apply_cfs(b_df)
bellwether0_s_cols[b_s_project] = b_s_cols
b_df = apply_smote(b_df)
b_y = b_df.Buggy
b_X = b_df.drop(labels=['Buggy'], axis=1)
b_clf_bellwether = LogisticRegression()
b_clf_bellwether.fit(b_X, b_y)
bellwether0_models[b_s_project] = b_clf_bellwether
else:
b_clf_bellwether = bellwether0_models[b_s_project]
b_s_cols = bellwether0_s_cols[b_s_project]
d_project = test_projects[i]
kf = StratifiedKFold(n_splits=2)
d_path = '/Users/suvodeepmajumder/Documents/AI4SE/bellwether_comminity/data/1385/converted/' + d_project
test_df = prepare_data(d_path)
test_df.reset_index(drop=True, inplace=True)
d = {'buggy': True, 'clean': False}
test_df['Buggy'] = test_df['Buggy'].map(d)
#test_df, x_s_cols = apply_cfs(test_df)
test_y = test_df.Buggy
test_X = test_df.drop(labels=['Buggy'], axis=1)
for train_index, test_index in kf.split(test_X, test_y):
X_train, X_test = test_X.iloc[train_index], test_X.iloc[
test_index]
y_train, y_test = test_y[train_index], test_y[test_index]
x_test_df = pd.concat([X_train, y_train], axis=1)
x_test_df = apply_smote(x_test_df)
y_train = x_test_df.Buggy
X_train = x_test_df.drop(labels=['Buggy'], axis=1)
clf_self = LogisticRegression()
clf_self.fit(X_train, y_train)
self_model[d_project] = clf_self
self_model_test[d_project] = [X_test, y_test]
_test_df = pd.concat(self_model_test[d_project], axis=1)
_df_test_loc = _test_df.LOC
_test_df_1 = copy.deepcopy(_test_df[s_cols])
_test_df_2 = copy.deepcopy(_test_df)
_test_df_3 = copy.deepcopy(_test_df[b_s_cols])
_test_df_4 = copy.deepcopy(_test_df[g_s_cols])
_test_df_5 = copy.deepcopy(_test_df)
y_test = _test_df_1.Buggy
X_test = _test_df_1.drop(labels=['Buggy'], axis=1)
predicted_bellwether = clf_bellwether.predict(X_test)
abcd = metrices.measures(y_test, predicted_bellwether,
_df_test_loc)
if 'f1' not in F.keys():
F['f1'] = []
F['precision'] = []
F['recall'] = []
F['g-score'] = []
F['d2h'] = []
F['pci_20'] = []
F['ifa'] = []
F['pd'] = []
F['pf'] = []
F['f1'].append(abcd.calculate_f1_score())
F['precision'].append(abcd.calculate_precision())
F['recall'].append(abcd.calculate_recall())
F['g-score'].append(abcd.get_g_score())
F['d2h'].append(abcd.calculate_d2h())
F['pci_20'].append(abcd.get_pci_20())
try:
F['ifa'].append(abcd.get_ifa_roc())
except:
F['ifa'].append(0)
F['pd'].append(abcd.get_pd())
F['pf'].append(abcd.get_pf())
try:
y_test = _test_df_2.Buggy
X_test = _test_df_2.drop(labels=['Buggy'], axis=1)
predicted_self = clf_self.predict(X_test)
abcd = metrices.measures(y_test, predicted_self,
_df_test_loc)
if 'f1' not in _F.keys():
_F['f1'] = []
_F['precision'] = []
_F['recall'] = []
_F['g-score'] = []
_F['d2h'] = []
_F['pci_20'] = []
_F['ifa'] = []
_F['pd'] = []
_F['pf'] = []
_F['f1'].append(abcd.calculate_f1_score())
_F['precision'].append(abcd.calculate_precision())
_F['recall'].append(abcd.calculate_recall())
_F['g-score'].append(abcd.get_g_score())
_F['d2h'].append(abcd.calculate_d2h())
_F['pci_20'].append(abcd.get_pci_20())
try:
_F['ifa'].append(abcd.get_ifa_roc())
except:
_F['ifa'].append(0)
_F['pd'].append(abcd.get_pd())
_F['pf'].append(abcd.get_pf())
except:
_F['f1'].append(0)
_F['precision'].append(0)
_F['recall'].append(0)
_F['g-score'].append(0)
_F['d2h'].append(0)
_F['pci_20'].append(0)
_F['ifa'].append(0)
_F['pd'].append(0)
_F['pf'].append(0)
b_y_test = _test_df_3.Buggy
b_X_test = _test_df_3.drop(labels=['Buggy'], axis=1)
predicted_bell0 = b_clf_bellwether.predict(b_X_test)
abcd = metrices.measures(b_y_test, predicted_bell0,
_df_test_loc)
if 'f1' not in b_F.keys():
b_F['f1'] = []
b_F['precision'] = []
b_F['recall'] = []
b_F['g-score'] = []
b_F['d2h'] = []
b_F['pci_20'] = []
b_F['ifa'] = []
b_F['pd'] = []
b_F['pf'] = []
b_F['f1'].append(abcd.calculate_f1_score())
b_F['precision'].append(abcd.calculate_precision())
b_F['recall'].append(abcd.calculate_recall())
b_F['g-score'].append(abcd.get_g_score())
b_F['d2h'].append(abcd.calculate_d2h())
b_F['pci_20'].append(abcd.get_pci_20())
try:
b_F['ifa'].append(abcd.get_ifa_roc())
except:
b_F['ifa'].append(0)
b_F['pd'].append(abcd.get_pd())
b_F['pf'].append(abcd.get_pf())
g_y_test = _test_df_4.Buggy
g_X_test = _test_df_4.drop(labels=['Buggy'], axis=1)
predicted_global = clf_global.predict(g_X_test)
abcd = metrices.measures(g_y_test, predicted_global,
_df_test_loc)
if 'f1' not in g_F.keys():
g_F['f1'] = []
g_F['precision'] = []
g_F['recall'] = []
g_F['g-score'] = []
g_F['d2h'] = []
g_F['pci_20'] = []
g_F['ifa'] = []
g_F['pd'] = []
g_F['pf'] = []
g_F['f1'].append(abcd.calculate_f1_score())
g_F['precision'].append(abcd.calculate_precision())
g_F['recall'].append(abcd.calculate_recall())
g_F['g-score'].append(abcd.get_g_score())
g_F['d2h'].append(abcd.calculate_d2h())
g_F['pci_20'].append(abcd.get_pci_20())
try:
g_F['ifa'].append(abcd.get_ifa_roc())
except:
g_F['ifa'].append(0)
g_F['pd'].append(abcd.get_pd())
g_F['pf'].append(abcd.get_pf())
r_y_test = _test_df_5.Buggy
r_X_test = _test_df_5.drop(labels=['Buggy'], axis=1)
_count_major = Counter(y_train)
_count_major = _count_major.most_common(1)[0][0]
predicted_random = [_count_major] * r_X_test.shape[0]
abcd = metrices.measures(r_y_test, predicted_random,
_df_test_loc)
if 'f1' not in r_F.keys():
r_F['f1'] = []
r_F['precision'] = []
r_F['recall'] = []
r_F['g-score'] = []
r_F['d2h'] = []
r_F['pci_20'] = []
r_F['ifa'] = []
r_F['pd'] = []
r_F['pf'] = []
r_F['f1'].append(abcd.calculate_f1_score())
r_F['precision'].append(abcd.calculate_precision())
r_F['recall'].append(abcd.calculate_recall())
r_F['g-score'].append(abcd.get_g_score())
r_F['d2h'].append(abcd.calculate_d2h())
r_F['pci_20'].append(abcd.get_pci_20())
try:
r_F['ifa'].append(abcd.get_ifa_roc())
except:
r_F['ifa'].append(0)
r_F['pd'].append(abcd.get_pd())
r_F['pf'].append(abcd.get_pf())
for goal in goals:
if goal == 'g':
_goal = 'g-score'
else:
_goal = goal
if _goal not in results.keys():
results[_goal] = {}
if d_project not in results[_goal].keys():
results[_goal][d_project] = []
results[_goal][d_project].append(np.median(F[_goal]))
results[_goal][d_project].append(np.median(b_F[_goal]))
results[_goal][d_project].append(np.median(_F[_goal]))
results[_goal][d_project].append(np.median(g_F[_goal]))
results[_goal][d_project].append(np.median(r_F[_goal]))
except Exception as e:
print(e)
continue
_cols = [
'level2_bellwether', 'default_bellwether', 'self', 'global', 'random',
'level1_bellwether', 'default_bellwether', 'self', 'global', 'random',
'level0_bellwether', 'default_bellwether', 'self', 'global', 'random'
]
print(results)
for key in results:
df = pd.DataFrame.from_dict(results[key],
orient='index',
columns=_cols)
if not Path(data_location).is_dir():
os.makedirs(Path(data_location))
df.to_csv(data_location + '/bellwether_' + key + '.csv')
return results
