def change_cwd_results(self):
cwd = os.getcwd()
os.chdir(
RESULT_PATH.format(reference=self.__reference_distribution,
data=self.full_data_label,
classes=self.full_classes_label))
return cwd
def load_data() -> pd.DataFrame:
all_files = []
for ref in SELECTED_REFERENCES:
for data_count_var in ['All Data', 'Partial Data']:
for class_count_var in ['Full Classes', 'Limited Classes']:
glob_files = list(filter(lambda z: '[' not in z, glob.glob(RESULT_PATH.format(reference=ref, data=data_count_var, classes=class_count_var) + '*.csv')))
for glob_file in glob_files:
assert glob_file not in all_files
all_files += glob_files
oa_df_list = []
for file in all_files:
ref_style, dim, data_style, hidden_layer, size, class_style = name_columns(file)
if dim == '3':
continue
temp = pd.read_csv(file)
if 'False Negatives' in temp.columns:
temp = dm.false_positive_rate(temp)
temp = dm.f1_score(temp)
temp = dm.specificity(temp)
temp = dm.false_omission_rate(temp)
temp.drop(columns=['True Positives', 'True Negatives', 'False Positives', 'False Negatives'], inplace=True)
columns = [ref_style + ' D ' + dim + ' ' + data_style + ' ' + col + ' HL ' + hidden_layer + ' ' + size + ' ' + class_style for col in temp.columns]
oa_df_list.append(temp.set_axis(columns, axis=1))
return pd.concat(oa_df_list, axis=1)
incomplete_knn_sizes = []
find_svm = False
incomplete_svm_sizes = []
for dim in [1, 2]:
print('Dimension:', dim)
for data in ['All Data', 'Partial Data']:
print('\t' + data)
for classes in ['Full Classes', 'Limited Classes']:
if classes == 'Limited Classes' and dim == 1:
continue
print('\t\t' + classes)
for reference in REFERENCE_LIST:
print('\t\t\t' + reference)
os.chdir(
RESULT_PATH.format(dim=dim,
reference=reference,
data=data,
classes=classes))
for size_var in SIZE_SET:
detail_size(size_var)
for dim in [1, 2]:
print('Dimension:', dim)
for reference in REFERENCE_LIST:
print('\t' + reference)
for data in ['All Data']:
print('\t\t' + data)
for classes in ['Full Classes', 'Limited Classes']:
print('\t\t\t' + classes + '\n\t\t\t\tAll Sizes', end='\t\t')
os.chdir(
RESULT_PATH.format(dim=dim,
reference=reference,
data=data,
elif class_count is None:
plt.title(data_count + ' ' + current_metric, fontsize=10)
plt.savefig('Same Data\\' + current_metric + '\\' + data_count + ' ' + current_title + '.png')
elif data_count is None:
plt.title(class_count + ' ' + current_metric, fontsize=10)
plt.savefig('Same Classes\\' + current_metric + '\\' + class_count + ' ' + current_title + '.png')
else:
plt.title(current_metric, fontsize=10)
plt.savefig(data_count + '\\' + class_count + '\\' + current_metric + '\\' + current_title + '.png')
plt.close('all')
x = range(1, NUM_EPOCHS + 1)
for class_count_var in ['Full Classes', 'Limited Classes']:
for data_count_var in ['All Data', 'Partial Data']:
os.chdir(RESULT_PATH.format(data=data_count_var, classes=class_count_var))
print(os.getcwd())
# Build DataFrame
oa_hidden_layer_list = []
for hidden_layer in NUM_HIDDEN_LAYERS:
hidden_layer_files = list(filter(lambda data: data_count_var not in data, glob.glob('PD*Hidden Layer ' + str(hidden_layer) + '*.csv')))
if len(hidden_layer_files) == 0:
continue
hidden_layer_files.sort(key=sort)
for file in hidden_layer_files:
size = file.split(',')[1].split(' ')[-1]
if int(size) < 100:
size = '0' + size
temp = pd.read_csv(file)
columns = [col + ' HL ' + str(hidden_layer) + ' S ' + size + '.' for col in temp.columns]
oa_hidden_layer_list.append(temp.set_axis(columns, axis=1))