def run_wilcoxon_test(df, name, folder): columns = df.columns.values t_df = pd.DataFrame(columns=columns, index=columns) for i in range(0, len(columns)-1): for j in range(i+1, len(columns)): stats, p_value = scipy.stats.mannwhitneyu( df[columns[i]].values, df[columns[j]].values, alternative='two-sided') t_df.iloc[j, i] = p_value write_df_to_csv(folder, t_df, name)
def do_ground_truth_all(files=None, kws=None):
print('\n> Defining ground truth for feature vector/s:')
[print("\t- " + os.path.relpath(file[0])) for file in files]
for p, n in files:
methods = get_fv_methods(p)
ground_truth = do_ground_truth(methods, kws)
write_df_to_csv(GT_DIR, gt_to_df(ground_truth, kws), n)
print('> Ground truth/s has/ve been written to folder "%s"' %
os.path.abspath(GT_DIR))
return GT_DIR
def compute_stats(metrics, classifiers, folder):
print(indent('\n- Computing metrics statistics ... '), end='')
stats = pd.DataFrame(columns=classifiers)
for key, val in metrics.items():
name = str(capitalize(key))
mean = val.mean(axis=0)
mean = mean.rename(name+' Mean')
stats = stats.append(mean)
median = val.median(axis=0)
median = median.rename(name+' Median')
stats = stats.append(median)
std = val.std(axis=0)
std = std.rename(name+' Standard Deviation')
stats = stats.append(std)
print('result:')
print(indent(stats.to_string(), spaces=10))
out = write_df_to_csv(folder, stats, 'stats')
print(indent('\n- Statistics written to file "%s"' % out))
def biased_clf_metrics_to_csv(labels, folder):
prec, rec, fscore, sup = precision_recall_fscore_support(labels, ones(len(labels)), average='binary')
df = pd.DataFrame({'precision': prec, 'recall': rec, 'fscore': fscore}, index=[0])
out = write_df_to_csv(folder, df, 'biased_metrics')
print(indent('\n- Biased classifier metrics ("precision", "recall" and "fscore") written to file "%s"' % out))
def do_all_cluster_from_path(
path=None,
target=None,
f=None,
n=5):
path = os.path.abspath(path)
target = os.path.abspath(target)
print('\n> Clustering God class methods in file/folder "%s"' % path)
# get all the feature vectors in the folder path
paths_and_names = get_paths_and_names(path)
for el in paths_and_names:
# applies teh function "f" (either k-means or hierarchical agglomerative)
df = f(el[0], n)
write_df_to_csv(target, df, el[1])
print('> Clusters have been written to folder "%s"' % target)
return target
def extract_feature_vectors(god_classes):
print('\n> Starting feature vector extraction...')
class_names = god_classes.class_name.tolist()
all_feat_vectors = {}
for src_path in god_classes.path_to_source.tolist():
# open the class source
with (open(src_path, 'r')) as jsc:
# parse the class
tree = jl.parse.parse(jsc.read())
# iterates through the file classes
for path, node in tree.filter(jl.parser.tree.ClassDeclaration):
if node.name in class_names: # check whether the class is a god class
# Generates the feature vector for each class
all_feat_vectors[node.name] = generate_all(node)
write_df_to_csv(FV_DIR, all_feat_vectors[node.name],
node.name)
fv_dir = os.path.abspath(FV_DIR)
print('> Feature vector/s has/ve been written to folder "%s"' % fv_dir)
return fv_dir
def label_feature_vectors(fv, fv_path, buggy_classes_dir):
print('\n> Creating labels for feature vector "%s"' % fv_path)
buggy_classes = get_buggy_classes(buggy_classes_dir)
label_feature_vector = get_label_feature_vector(fv, buggy_classes)
path = write_df_to_csv(
DEF_LFV_DIR, label_feature_vector,
gen_name_with_suffix('label_feature_vector',
get_dir_time_suffix(fv_path, 'feature_vector')))
print('> Labeled feature vector has been written to file "%s"' %
os.path.abspath(path))
return label_feature_vector, path
def run_training(classifier, classifier_name, fv_path, tt, r_num):
l_precision, l_recall, l_fscore, l_accurancy = [], [], [], []
print(indent('\n- Training classifier "%s"...' % classifier_name))
for i in range(0, len(r_num)):
pred, acc = run_classifier(
classifier,
tt['x_trains'][i],
tt['x_tests'][i],
tt['y_trains'][i],
tt['y_tests'][i])
prec, rec, f1 = get_prec_recall_fscore(tt['y_tests'][i], pred)
l_accurancy.append(acc)
l_precision.append(prec)
l_recall.append(rec)
l_fscore.append(f1)
df = pd.DataFrame(
{
"r_num": r_num,
"accuracy": l_accurancy,
"precision": l_precision,
"recall": l_recall,
"fscore": l_fscore
})
tr_folder = DEF_TR_DIR + '/' + get_dir_time_suffix(fv_path, 'label_feature_vector-')
path = write_df_to_csv(tr_folder, df, classifier_name.replace(' ', ''))
print(indent('\nResults written to file "%s"' % path, spaces=10))
print_averages(df)
make_plot(r_num, df[['accuracy', 'precision', 'recall', 'fscore']], classifier_name, tr_folder)
def extract_feature_vectors(root):
print('\n> Starting feature vector extraction for project "%s"' % root)
df = pd.DataFrame(columns=FV_COLS)
for t_class in get_top_classes(root):
mth, fld, rfc, ints = get_class_metrics(t_class)
sz, cpx, ex, ret = get_methods_metrics(t_class)
bcm, nml, wrd, dcm = get_npl_metrics(t_class)
df = df.append(
{
'class': t_class.name, # class name
'MTH': mth,
'FLD': fld,
'RFC': rfc,
'INT': ints, # CLASS METRICS
'SZ': sz,
'CPX': cpx,
'EX': ex,
'RET': ret, # METHOD METRICS
'BCM': bcm,
'NML': nml,
'WRD': wrd,
'DCM': dcm # NPL METRICS
},
ignore_index=True,
sort=-1)
df = df_sort_cols(df, FV_COLS)
path = write_df_to_csv(DEF_FV_DIR, df,
gen_name_with_time('feature_vector'))
print('> Feature vector/s has/ve been written to file "%s"' %
os.path.abspath(path))
return df, path