def tnb(source, target, n_rep=12):
"""
TNB: Transfer Naive Bayes
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
plot_data =[("Xalan", "Log4j", "Lucene", "Poi", "Velocity")]
for tgt_name, tgt_path in target.iteritems():
stats = []
charts = []
print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
# print("{} \r".format(src_name[0].upper() + src_name[1:]))
src = list2dataframe(src_path.data)
tgt = list2dataframe(tgt_path.data)
pd, pf, g, auc = [], [], [], []
for _ in xrange(n_rep):
lo, hi, test_mass = target_details(tgt)
weights = get_weights(maxs=hi, mins=lo, train_set=src, test_set=tgt)
_train, __test = weight_training(weights=weights, training_instance=src, test_instance=tgt)
actual, predicted, distribution = predict_defects(train=_train, test=__test)
# loc = tgt["$loc"].values
# loc = loc * 100 / np.max(loc)
# recall, loc, au_roc = get_curve(loc, actual, predicted, distribution)
# effort_plot(recall, loc,
# save_dest=os.path.abspath(os.path.join(root, "plot", "plots", tgt_name)),
# save_name=src_name)
p_d, p_f, p_r, rc, f_1, e_d, _g, auroc = abcd(actual, predicted, distribution, threshold=0.4)
pd.append(p_d)
pf.append(p_f)
g.append(_g)
auc.append(int(auroc))
stats.append([src_name, int(np.mean(pd)), int(np.std(pd)),
int(np.mean(pf)), int(np.std(pf)),
int(np.mean(auc)), int(np.std(auc))]) # ,
# int(np.mean(g)), int(np.std(g))])
# print("")
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[-2], reverse=True), # Sort by G Score
columns=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"]) # ,
# "G (Mean)", "G (Std)"])
print(tabulate(stats,
headers=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
return result
def tca_plus(source, target, verbose=False, n_rep=12):
"""
TCA: Transfer Component Analysis
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
for tgt_name, tgt_path in target.iteritems():
stats = []
print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
src = list2dataframe(src_path)
tgt = list2dataframe(tgt_path)
pd, pf, pr, f1, g, auc = [], [], [], [], [], []
dcv_src, dcv_tgt = get_dcv(src, tgt)
for _ in xrange(n_rep):
norm_src, norm_tgt = smart_norm(src, tgt, dcv_src, dcv_tgt)
_train, __test = map_transform(norm_src, norm_tgt)
actual, predicted, distribution = predict_defects(
train=_train, test=__test)
p_d, p_f, p_r, rc, f_1, e_d, _g, auroc = abcd(
actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
pr.append(p_f)
f1.append(p_f)
g.append(_g)
auc.append(int(auroc))
stats.append([
src_name,
int(np.mean(pd)),
int(np.std(pd)),
int(np.mean(pf)),
int(np.std(pf)),
int(np.mean(auc)),
int(np.std(auc))
]) # ,
stats = pandas.DataFrame(
sorted(stats, key=lambda lst: lst[0]), # Sort by G Score
columns=["Name", "Pd", "Pf", "Prec", "F1", "G", "AUC"]) # ,
if verbose:
print(
tabulate(
stats,
headers=["Name", "Pd", "Pf", "Prec", "F1", "G", "AUC"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
return result
def tca_plus(source, target, verbose=False, n_rep=12):
"""
TCA: Transfer Component Analysis
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
for tgt_name, tgt_path in target.iteritems():
stats = []
print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
# set_trace()
src = list2dataframe(src_path)
tgt = list2dataframe(tgt_path)
# set_trace()
pd, pf, g, auc = [], [], [], []
dcv_src, dcv_tgt = get_dcv(src, tgt)
for _ in xrange(n_rep):
recall, loc = None, None
norm_src, norm_tgt = smart_norm(src, tgt, dcv_src, dcv_tgt)
_train, __test = map_transform(norm_src, norm_tgt)
# for k in np.arange(0.1,1,0.1):
actual, predicted, distribution = predict_defects(train=_train, test=__test)
# loc = tgt["$loc"].values
# loc = loc * 100 / np.max(loc)
# recall, loc, au_roc = get_curve(loc, actual, predicted)
# effort_plot(recall, loc,
# save_dest=os.path.abspath(os.path.join(root, "plot", "plots", tgt_name)),
# save_name=src_name)
p_d, p_f, p_r, rc, f_1, e_d, _g, _ = abcd(actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
g.append(_g)
# auc.append(int(au_roc))
# set_trace()
stats.append([src_name, int(np.mean(pd)), int(np.std(pd)),
int(np.mean(pf)), int(np.std(pf))]) # ,
# int(np.mean(auc)), int(np.std(auc))]) # ,
# int(np.mean(g)), int(np.std(g))])
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[0]), # Sort by G Score
columns=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)"]) # ,,
# "AUC (Mean)", "AUC (Std)"]) # ,
# "G (Mean)", "G (Std)"])
result.update({tgt_name: stats})
# set_trace()
return result
def seer(source, target, n_rep=20, n_redo=5):
"""
seer: Causal Inference Learning
:param source:
:param target:
:return: result: A dictionary of estimated
"""
result = dict()
t0 = time()
for tgt_name, tgt_path in target.iteritems():
stats = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
src = list2dataframe(src_path.data)
tgt = list2dataframe(tgt_path.data)
pd, pf, g = [], [], []
matched_src = metrics_match(src, tgt, n_redo)
for n in xrange(n_rep):
target_columns = []
source_columns = []
all_columns = [(key, val[0], val[1]) for key, val in matched_src.iteritems() if val[1] > 1]
all_columns = sorted(all_columns, key=lambda x: x[-1], reverse=True) # Sort descending
# Filter all columns to remove dupes
for elem in all_columns:
if not elem[1] in source_columns:
target_columns.append(elem[0])
source_columns.append(elem[1])
_train, __test = src[source_columns + [src.columns[-1]]], \
tgt[target_columns + [tgt.columns[-1]]]
# _train, __test = map_transform(src[source_columns + [src.columns[-1]]],
# tgt[target_columns + [tgt.columns[-1]]])
# set_trace()
actual, predicted = predict_defects(train=_train, test=__test)
p_d, p_f, p_r, rc, f_1, e_d, _g = abcd(actual, predicted)
pd.append(p_d)
pf.append(p_f)
g.append(e_d)
stats.append([src_name, round(np.mean(pd), 2), round(np.std(pd)),
round(np.mean(pf), 2), round(np.std(pf), 2),
round(np.mean(g), 2), round(np.std(g), 2)])
# set_trace()
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[0]), # Sort by G Score
columns=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"G (Mean)", "G (Std)"])
set_trace()
result.update({tgt_name: stats})
return result
def tca_plus(source, target, n_rep=12):
"""
TCA: Transfer Component Analysis
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
for tgt_name, tgt_path in target.iteritems():
stats = []
print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
print("{} \r".format(src_name[0].upper() + src_name[1:]))
src = list2dataframe(src_path.data)
tgt = list2dataframe(tgt_path.data)
pd, pf, g, auc = [], [], [], []
dcv_src, dcv_tgt = get_dcv(src, tgt)
for _ in xrange(n_rep):
recall, loc = None, None
norm_src, norm_tgt = smart_norm(src, tgt, dcv_src, dcv_tgt)
_train, __test = map_transform(norm_src, norm_tgt)
try:
actual, predicted, distribution = predict_defects(train=_train, test=__test)
except:
set_trace()
p_d, p_f, p_r, rc, f_1, e_d, _g, auroc = abcd(actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
g.append(_g)
auc.append(int(auroc))
stats.append([src_name, int(np.mean(pd)), int(np.std(pd)),
int(np.mean(pf)), int(np.std(pf)),
int(np.mean(auc)), int(np.std(auc))]) # ,
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[-2], reverse=True), # Sort by G Score
columns=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"]) # ,
print(tabulate(stats,
headers=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
return result
def vcb(source, target, n_rep=12):
"""
TNB: Transfer Naive Bayes
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
plot_data = [("Xalan", "Log4j", "Lucene", "Poi", "Velocity")]
for tgt_name, tgt_path in target.iteritems():
stats = []
charts = []
print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
# print("{} \r".format(src_name[0].upper() + src_name[1:]))
src = list2dataframe(src_path.data)
tgt = list2dataframe(tgt_path.data)
pd, pf, g, auc = [], [], [], []
for _ in xrange(n_rep):
_train, clf_w, classifiers = weight_training(train=src, test=tgt)
actual, predicted, distribution = predict_defects(tgt, clf_w, classifiers)
loc = tgt["$loc"].values
loc = loc * 100 / np.max(loc)
recall, loc, au_roc = get_curve(loc, actual, predicted, distribution)
effort_plot(recall, loc,
save_dest=os.path.abspath(os.path.join(root, "plot", "plots", tgt_name)),
save_name=src_name)
p_d, p_f, p_r, rc, f_1, e_d, _g, auroc = abcd(actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
g.append(_g)
auc.append(int(auroc))
stats.append([src_name, int(np.mean(pd)), int(np.std(pd)),
int(np.mean(pf)), int(np.std(pf)),
int(np.mean(auc)), int(np.std(auc))])
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[-2], reverse=True), # Sort by G Score
columns=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"]) # ,
# "G (Mean)", "G (Std)"])
print(tabulate(stats,
headers=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
return result
def tca_plus(source, target, verbose=True, n_rep=12):
"""
TCA: Transfer Component Analysis
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
for tgt_name, tgt_path in target.iteritems():
stats = []
if verbose: print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
src = pandas.read_csv(src_path)
tgt = pandas.read_csv(tgt_path)
pd, pf, pr, f1, g, auc = [], [], [], [], [], []
dcv_src, dcv_tgt = get_dcv(src, tgt)
for _ in xrange(n_rep):
norm_src, norm_tgt = smart_norm(src, tgt, dcv_src, dcv_tgt)
_train, __test = map_transform(norm_src.dropna(axis=1, inplace=False),
norm_tgt.dropna(axis=1, inplace=False))
actual, predicted, distribution = predict_defects(train=_train, test=__test)
p_d, p_f, p_r, rc, f_1, e_d, _g, auroc = abcd(actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
pr.append(p_r)
f1.append(f_1)
g.append(_g)
auc.append(int(auroc))
stats.append([src_name, int(np.mean(pd)), int(np.mean(pf)),
int(np.mean(pr)), int(np.mean(f1)),
int(np.mean(g)), int(np.mean(auc))]) # ,
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[-2], reverse=True), # Sort by G Score
columns=["Name", "Pd", "Pf", "Prec", "F1", "G", "AUC"]) # ,
if verbose: print(tabulate(stats,
headers=["Name", "Pd", "Pf", "Prec", "F1", "G", "AUC"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
return result
def bellw(source, target, n_rep=12, verbose=False):
"""
TNB: Transfer Naive Bayes
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
for tgt_name, tgt_path in target.iteritems():
stats = []
charts = []
print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
src = list2dataframe(src_path.data)
tgt = list2dataframe(tgt_path.data)
pd, pf, pr, f1, g, auc = [], [], [], [], [], []
for _ in xrange(n_rep):
_train, __test = weight_training(test_instance=tgt, training_instance=src)
actual, predicted, distribution = predict_defects(train=_train, test=__test)
p_d, p_f, p_r, rc, f_1, e_d, _g, auroc = abcd(actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
pr.append(p_r)
f1.append(f_1)
g.append(_g)
auc.append(int(auroc))
stats.append([src_name, int(np.mean(pd)), int(np.mean(pf)),
int(np.mean(pr)), int(np.mean(f1)),
int(np.mean(g)), int(np.mean(auc))]) # ,
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[-2], reverse=True), # Sort by G Score
columns=["Name", "Pd", "Pf", "Prec", "F1", "G", "AUC"]) # ,
if verbose: print(tabulate(stats,
headers=["Name", "Pd", "Pf", "Prec", "F1", "G", "AUC"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
return result
def tca_plus(source, target, n_rep=12):
"""
TCA: Transfer Component Analysis
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
for tgt_name, tgt_path in target.iteritems():
stats = []
print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
# set_trace()
src = list2dataframe(src_path)
tgt = list2dataframe(tgt_path)
# set_trace()
pd, pf, g, auc = [], [], [], []
dcv_src, dcv_tgt = get_dcv(src, tgt)
for _ in xrange(n_rep):
recall, loc = None, None
norm_src, norm_tgt = smart_norm(src, tgt, dcv_src, dcv_tgt)
_train, __test = map_transform(norm_src, norm_tgt)
# for k in np.arange(0.1,1,0.1):
actual, predicted, distribution = predict_defects(train=_train, test=__test)
# loc = tgt["$loc"].values
# loc = loc * 100 / np.max(loc)
# recall, loc, au_roc = get_curve(loc, actual, predicted)
# effort_plot(recall, loc,
# save_dest=os.path.abspath(os.path.join(root, "plot", "plots", tgt_name)),
# save_name=src_name)
p_d, p_f, p_r, rc, f_1, e_d, _g, _ = abcd(actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
g.append(_g)
# auc.append(int(au_roc))
# set_trace()
stats.append([src_name, int(np.mean(pd)), int(np.std(pd)),
int(np.mean(pf)), int(np.std(pf)),
int(np.mean(auc)), int(np.std(auc))]) # ,
# int(np.mean(g)), int(np.std(g))])
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[0]), # Sort by G Score
columns=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"]) # ,
# "G (Mean)", "G (Std)"])
print(tabulate(stats,
headers=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
# set_trace()
return result
def tnb(source, target, verbose=False, n_rep=12):
"""
TNB: Transfer Naive Bayes
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
for tgt_name, tgt_path in target.iteritems():
stats = []
if verbose: print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
src = pandas.read_csv(src_path)
tgt = pandas.read_csv(tgt_path)
pd, pf, pr, f1, g, auc = [], [], [], [], [], []
for _ in xrange(n_rep):
lo, hi, test_mass = target_details(tgt)
weights = get_weights(maxs=hi,
mins=lo,
train_set=src,
test_set=tgt)
_train, __test = weight_training(weights=weights,
training_instance=src,
test_instance=tgt)
actual, predicted, distribution = predict_defects(
train=_train, test=__test)
p_d, p_f, p_r, rc, f_1, e_d, _g, auroc = abcd(
actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
pr.append(p_r)
f1.append(f_1)
g.append(_g)
auc.append(int(auroc))
stats.append([
src_name,
int(np.mean(pd)),
int(np.mean(pf)),
int(np.mean(pr)),
int(np.mean(f1)),
int(np.mean(g)),
int(np.mean(auc))
]) # ,
stats = pandas.DataFrame(
sorted(stats, key=lambda lst: lst[-2],
reverse=True), # Sort by G Score
columns=["Name", "Pd", "Pf", "Prec", "F1", "G", "AUC"]) # ,
if verbose:
print(
tabulate(
stats,
headers=["Name", "Pd", "Pf", "Prec", "F1", "G", "AUC"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
return result
def seer(source, target, n_rep=20, n_redo=5):
"""
seer: Causal Inference Learning
:param source:
:param target:
:return: result: A dictionary of estimated
"""
result = dict()
t0 = time()
for tgt_name, tgt_path in target.iteritems():
stats = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
src = list2dataframe(src_path.data)
tgt = list2dataframe(tgt_path.data)
pd, pf, g = [], [], []
matched_src = metrics_match(src, tgt, n_redo)
for n in xrange(n_rep):
target_columns = []
source_columns = []
all_columns = [(key, val[0], val[1])
for key, val in matched_src.iteritems()
if val[1] > 1]
all_columns = sorted(all_columns,
key=lambda x: x[-1],
reverse=True) # Sort descending
# Filter all columns to remove dupes
for elem in all_columns:
if not elem[1] in source_columns:
target_columns.append(elem[0])
source_columns.append(elem[1])
_train, __test = src[source_columns + [src.columns[-1]]], \
tgt[target_columns + [tgt.columns[-1]]]
# _train, __test = map_transform(src[source_columns + [src.columns[-1]]],
# tgt[target_columns + [tgt.columns[-1]]])
# set_trace()
actual, predicted = predict_defects(train=_train,
test=__test)
p_d, p_f, p_r, rc, f_1, e_d, _g = abcd(actual, predicted)
pd.append(p_d)
pf.append(p_f)
g.append(e_d)
stats.append([
src_name,
round(np.mean(pd), 2),
round(np.std(pd)),
round(np.mean(pf), 2),
round(np.std(pf), 2),
round(np.mean(g), 2),
round(np.std(g), 2)
])
# set_trace()
stats = pandas.DataFrame(
sorted(stats, key=lambda lst: lst[0]), # Sort by G Score
columns=[
"Name", "Pd (Mean)", "Pd (Std)", "Pf (Mean)", "Pf (Std)",
"G (Mean)", "G (Std)"
])
set_trace()
result.update({tgt_name: stats})
return result
def seer(source, target, n_rep=20, n_redo=5):
"""
seer: Causal Inference Learning
:param source:
:param target:
:return: result: A dictionary of estimated
"""
result = dict()
t0 = time()
for tgt_name, tgt_path in target.iteritems():
stats = []
print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
src = list2dataframe(src_path.data)
tgt = list2dataframe(tgt_path.data)
pd, pf, g, auc = [], [], [], []
matched_src = metrics_match(src, tgt, n_redo)
for n in xrange(n_rep):
target_columns = []
source_columns = []
all_columns = [(key, val[0], val[1]) for key, val in matched_src.iteritems() if val[1] > 1]
all_columns = sorted(all_columns, key=lambda x: x[-1], reverse=True) # Sort descending
# Filter all columns to remove dupes
for elem in all_columns:
if not elem[1] in source_columns:
target_columns.append(elem[0])
source_columns.append(elem[1])
selected_col = list(set(target_columns).intersection(source_columns))
_train, __test = map_transform(src[selected_col + [src.columns[-1]]],
tgt[selected_col + [tgt.columns[-1]]])
# _train, __test = src[source_columns + [src.columns[-1]]], \
# tgt[target_columns + [tgt.columns[-1]]]
# set_trace()
actual, predicted, distribution = predict_defects(train=_train, test=__test)
p_d, p_f, p_r, rc, f_1, e_d, _g, auroc = abcd(actual, predicted, distribution)
pd.append(p_d)
pf.append(p_f)
g.append(e_d)
auc.append(int(auroc))
stats.append([src_name, int(np.mean(pd)), int(np.std(pd)),
int(np.mean(pf)), int(np.std(pf)),
int(np.mean(auc)), int(np.std(auc))]) # ,
stats = pandas.DataFrame(sorted(stats, key=lambda lst: lst[-2], reverse=True), # Sort by G Score
columns=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"]) # ,
# "G (Mean)", "G (Std)"])
print(tabulate(stats,
headers=["Name", "Pd (Mean)", "Pd (Std)",
"Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
return result
def tnb(source, target, verbose=False, n_rep=12):
"""
TNB: Transfer Naive Bayes
:param source:
:param target:
:param n_rep: number of repeats
:return: result
"""
result = dict()
plot_data = [("Xalan", "Log4j", "Lucene", "Poi", "Velocity")]
for tgt_name, tgt_path in target.iteritems():
stats = []
charts = []
if verbose: print("{} \r".format(tgt_name[0].upper() + tgt_name[1:]))
val = []
for src_name, src_path in source.iteritems():
if not src_name == tgt_name:
# print("{} \r".format(src_name[0].upper() + src_name[1:]))
src = list2dataframe(src_path.data)
tgt = list2dataframe(tgt_path.data)
pd, pf, g, auc = [], [], [], []
lo, hi, test_mass = target_details(tgt)
weights = get_weights(maxs=hi,
mins=lo,
train_set=src,
test_set=tgt)
_train = weight_training(weights=weights,
training_instance=src)
__test = (tgt[tgt.columns[:-1]] - tgt[tgt.columns[:-1]].min()
) / (tgt[tgt.columns[:-1]].max() -
tgt[tgt.columns[:-1]].min())
__test[tgt.columns[-1]] = tgt[tgt.columns[-1]]
actual, predicted, distribution = predict_defects(train=_train,
test=__test)
loc = tgt["$loc"].values
loc = loc * 100 / np.max(loc)
recall, loc, au_roc = get_curve(loc, actual, predicted)
effort_plot(recall,
loc,
save_dest=os.path.abspath(
os.path.join(root, "plot", "plots", tgt_name)),
save_name=src_name)
p_d, p_f, p_r, rc, f_1, e_d, _g = abcd(actual, predicted,
distribution)
pd.append(p_d)
pf.append(p_f)
g.append(_g)
auc.append(int(au_roc))
stats.append([
src_name,
int(np.mean(pd)),
int(np.std(pd)),
int(np.mean(pf)),
int(np.std(pf)),
int(np.mean(auc)),
int(np.std(auc))
]) # ,
# int(np.mean(g)), int(np.std(g))])
# print("")
stats = pandas.DataFrame(
sorted(stats, key=lambda lst: lst[0]), # Sort by G Score
columns=[
"Name", "Pd (Mean)", "Pd (Std)", "Pf (Mean)", "Pf (Std)",
"AUC (Mean)", "AUC (Std)"
]) # ,
# "G (Mean)", "G (Std)"])
if verbose:
print(
tabulate(stats,
headers=[
"Name", "Pd (Mean)", "Pd (Std)", "Pf (Mean)",
"Pf (Std)", "AUC (Mean)", "AUC (Std)"
],
showindex="never",
tablefmt="fancy_grid"))
result.update({tgt_name: stats})
set_trace()
return result