ratio, monotone,
votes_params['miss_data_symbol'],
votes_params['mnar_values'])
else:
pert_data = votes_train
path = os.path.join(
perturb_folder,
'{}_train_pert_mono_{}_ratio_{}.csv'.format(dataname, monotone, ratio))
# save perturbed data to disk as csv
print '\tSaving perturbed data to {}'.format(path)
np.savetxt(path, pert_data, delimiter=",", fmt="%s")
# impute data given imp_methods in params.py
for imp_method in votes_params['imp_methods']:
print '\tImputing with {}'.format(imp_method)
imp = Imputer()
data = impute(pert_data, imp, imp_method, votes_params)
path = "data/imputed/{}_{}_mono_{}_ratio_{}.csv".format(
dataname, imp_method, monotone, ratio)
# save data as csv
print '\tSaving imputed data to {}'.format(path)
np.savetxt(path, data, delimiter=",", fmt="%s")
# binarize data
data_scaled_bin = imp.binarize_data(data, votes_params['cat_cols'],
votes_params['miss_data_symbol'])
# convert to float
data_scaled_bin = data_scaled_bin.astype(float)
# add labels as last column
data_scaled_bin = np.hstack((data_scaled_bin, labels_train))
pert_data = x
print "\tRatio is {} of {}".format(
np.sum(pert_data == adult_params['miss_data_symbol']),
len(pert_data) * len(adult_params['cat_cols']))
path = os.path.join(
perturb_folder,
'adult_train_pert_mono_{}_ratio_{}.csv'.format(monotone, ratio))
# save perturbed data to disk as csv
print '\tSaving perturbed data to {}'.format(path)
np.savetxt(path, pert_data, delimiter=",", fmt="%s")
for imp_method in adult_params['imp_methods']:
print '\tImputing with {}'.format(imp_method)
imp = Imputer()
data = impute(pert_data, imp, imp_method, adult_params)
path = "data/imputed/{}_mono_{}_ratio_{}.csv".format(
imp_method, monotone, ratio)
# save data as csv
print '\tSaving imputed data to {}'.format(path)
np.savetxt(path, data, delimiter=",", fmt="%s")
# scale continuous variables and convert categorial to one-hot
# store the scaler objects to be used on the test set
scaler_path = os.path.join(scalers_folder,
"{}_scaler".format(imp_method))
if os.path.isfile(scaler_path):
scaler_dict = pkl.load(open(scaler_path, "rb"))
else:
votes_train, votes_params['cat_cols'], ratio, monotone,
votes_params['miss_data_symbol'], votes_params['mnar_values'])
else:
pert_data = votes_train
path = os.path.join(perturb_folder,
'{}_train_pert_mono_{}_ratio_{}.csv'.format(dataname,
monotone,
ratio))
# save perturbed data to disk as csv
print '\tSaving perturbed data to {}'.format(path)
np.savetxt(path, pert_data, delimiter=",", fmt="%s")
# impute data given imp_methods in params.py
for imp_method in votes_params['imp_methods']:
print '\tImputing with {}'.format(imp_method)
imp = Imputer()
data = impute(pert_data, imp, imp_method, votes_params)
path = "data/imputed/{}_{}_mono_{}_ratio_{}.csv".format(dataname,
imp_method,
monotone,
ratio)
# save data as csv
print '\tSaving imputed data to {}'.format(path)
np.savetxt(path, data, delimiter=",", fmt="%s")
# binarize data
data_scaled_bin = imp.binarize_data(data,
votes_params['cat_cols'],
votes_params['miss_data_symbol'])
# convert to float
data_scaled_bin = data_scaled_bin.astype(float)
pert_data = x
print "\tRatio is {} of {}".format(
np.sum(pert_data == adult_params['miss_data_symbol']),
len(pert_data) * len(adult_params['cat_cols']))
path = os.path.join(perturb_folder,
'adult_train_pert_mono_{}_ratio_{}.csv'.format(monotone,
ratio))
# save perturbed data to disk as csv
print '\tSaving perturbed data to {}'.format(path)
np.savetxt(path, pert_data, delimiter=",", fmt="%s")
for imp_method in adult_params['imp_methods']:
print '\tImputing with {}'.format(imp_method)
imp = Imputer()
data = impute(pert_data, imp, imp_method, adult_params)
path = "data/imputed/{}_mono_{}_ratio_{}.csv".format(imp_method,
monotone,
ratio)
# save data as csv
print '\tSaving imputed data to {}'.format(path)
np.savetxt(path, data, delimiter=",", fmt="%s")
# scale continuous variables and convert categorial to one-hot
# store the scaler objects to be used on the test set
scaler_path = os.path.join(scalers_folder,
"{}_scaler".format(imp_method))
if os.path.isfile(scaler_path):
scaler_dict = pkl.load(open(scaler_path, "rb"))
# binarize labels
labels = (np.array(x[:, -1]) == '>50K').astype(int)
labels = labels.reshape((-1, 1))
# save labels in binary and one-hot representations
labels.dump(os.path.join(labels_test_folder, 'labels_bin_test.np'))
# remove redundant education-number and labels features
x = delete(x, (4, 14), 1)
# instantiate Imputer
imp = Imputer()
for imp_method in adult_params['imp_methods']:
print 'Imputing with {}'.format(imp_method)
data = impute(x, imp, imp_method, adult_params)
# load respective scaler
scaler_path = os.path.join(scalers_folder,
"{}_scaler".format(imp_method))
scaler_dict = pickle.load(open(scaler_path, "rb"))
for name, scaler in scaler_dict.items():
print 'Scaling with {}'.format(name)
# scale and binarize, adding one col for missing value in all categ vars
data_scaled = np.copy(data)
data_scaled[:, adult_params['non_cat_cols']] = scaler.transform(
data[:, adult_params['non_cat_cols']].astype(float))
data_scaled_bin = imp.binarize_data(data_scaled,
adult_params['cat_cols'],
adult_params['miss_data_symbol'])
