def main():
opt = parse_arg()
print('')
print(opt)
x_batches, y_batches, img_count = func.prep_data(opt.currency,
opt.days_ahead)
trainset_full, testset = func.test_data(x_batches, y_batches, opt.split,
img_count)
func.change_dir(opt.CNN_only, opt.currency, opt.days_ahead)
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
model = func.prep_model(opt.CNN_only,
n_tree=opt.n_tree,
tree_depth=opt.tree_depth,
tree_feature_rate=opt.tree_feature_rate)
model = model.to(device)
optim = func.prep_optim(model, opt.optimizer)
train_loss, val_loss, class_err, roc, auc, conf_mat = func.training(
model, opt.CNN_only, opt.batch_size, opt.epochs, device, optim,
trainset_full, testset)
accuracy = 1 - class_err
print("accuracy on testset", round(100 * accuracy), "%")
print("AUC testset", round(auc, 3))
print("confusion matrix", conf_mat)
test_results = dict(auc=auc, accuracy=accuracy)
with open('test_results.csv', 'w') as f:
w = csv.DictWriter(f, test_results.keys())
w.writeheader()
w.writerow(test_results)
with open('conf_mat.csv', 'w') as f:
w = csv.DictWriter(f, conf_mat.keys())
w.writeheader()
w.writerow(conf_mat)
tpr = roc[0]
fpr = roc[1]
roc_curve = [tpr, fpr]
with open("roc_curve.csv", 'w') as f:
writer = csv.writer(f)
writer.writerows(roc_curve)
# Print the console output
sys.stout = open("arguments.txt", "w")
print(opt, file=sys.stout)
import glob
import os
import functions as func
def parse_arg():
parser = argparse.ArgumentParser()
parser.add_argument('-currency', choices=['EURUSD', 'USDJPY', 'GBPUSD'], type=str, default="EURUSD")
parser.add_argument('-CNN_only', action='store_true')
parser.add_argument('-days_ahead', type=int, default=1)
opt = parser.parse_args()
return opt
opt = parse_arg()
func.change_dir(opt.CNN_only, opt.currency, opt.days_ahead)
path = os.getcwd()
all_results = glob.glob(path + '/*.csv')
try:
all_results.remove(path + '/val_results_full.csv')
except:
pass
try:
all_results.remove(path + '/roc_curve.csv')
except:
pass
try:
name_of_dir = input("Введите имя папки: ")
os.mkdir(name_of_dir)
elif choice == '2':
functions.remove_file_or_dir()
elif choice == '3':
functions.copy_file_or_dir()
elif choice == '4':
cwd = os.getcwd()
print(os.listdir(path=cwd))
elif choice == '5':
print(functions.print_dirs())
elif choice == '6':
print(functions.print_files())
elif choice == '7':
print(platform.uname())
elif choice == '8':
print(functions.info_about_creator())
elif choice == '9':
victory()
elif choice == '10':
account_f()
elif choice == '11':
new_path = input("Введите путь к желаемой директории: ")
functions.change_dir(new_path)
elif choice == '12':
functions.save_fails_and_dirs()
elif choice == '13':
functions.exit()
break
else:
print('Неверный пункт меню')
functions.display()
choice = input('Выберите пункт меню')
if choice == '1':
name_of_dir = input("Введите имя папки: ")
os.mkdir(name_of_dir)
elif choice == '2':
functions.remove_file_or_dir()
elif choice == '3':
functions.copy_file_or_dir()
elif choice == '4':
cwd = os.getcwd()
print(os.listdir(path=cwd))
elif choice == '5':
functions.print_dirs()
elif choice == '6':
functions.print_files()
elif choice == '7':
print(platform.uname())
elif choice == '8':
functions.info_about_creator()
elif choice == '9':
victory()
elif choice == '10':
account()
elif choice == '11':
functions.change_dir()
elif choice == '12':
exit()
break
else:
print('Неверный пункт меню')
elif choice == '6':
functions.list_file() # файлы рабочей дирректории
elif choice == '7':
functions.list_folders() # папки рабочей дирректории
elif choice == '8':
print('Информация об ОС:')
print(platform.uname()) # информация об ОС
elif choice == '9':
print('Создатель программы: Шаров Игорь') # автор программы
elif choice == '10': # викторина
print('Викторина')
victory.victory()
elif choice == '11': # банковский счет
print('Мой банковский счет. Меню:')
use_functions.banks()
elif choice == '11':
functions.change_dir() # смена рабочей дирректории
elif choice == '0':
print('Вы вышли из программы')
break
else:
print('Вы ввели неверный пункт меню')
def test_change_dir():
old_onlydirs = functions.print_dirs()
functions.change_dir("C:/Users/Ольга")
new_onlydirs = functions.print_dirs()
assert old_onlydirs != new_onlydirs
def main():
opt = parse_arg()
print('')
print('CNN_only:', opt.CNN_only, ', currency:', opt.currency,
', train/valset split:', opt.split)
x_batches, y_batches, img_count = func.prep_data(opt.currency,
opt.days_ahead)
trainset_val, valset = func.validation_data(x_batches, y_batches,
opt.split, img_count)
func.change_dir(opt.CNN_only, opt.currency, opt.days_ahead)
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
grid = func.prep_grid(opt.CNN_only, opt.grid_nr)
print('')
print("testing", grid.shape[0],
"hyperparameter combinations in gridsearch")
results = []
for i in range(len(grid)):
hyparams = grid.loc[i]
print('')
print(grid.loc[[i]])
batch_size = int(hyparams['batch_size'])
epochs = int(hyparams['epochs'])
optimizer = str(hyparams['optimizer'])
if opt.CNN_only:
n_tree = 'n/a'
tree_depth = 'n/a'
tree_feature_rate = 'n/a'
else:
n_tree = int(hyparams['n_tree'])
tree_depth = int(hyparams['tree_depth'])
tree_feature_rate = float(hyparams['feature_rate'])
model = func.prep_model(opt.CNN_only,
n_tree=n_tree,
tree_depth=tree_depth,
tree_feature_rate=tree_feature_rate)
model = model.to(device)
optim = func.prep_optim(model, optimizer)
train_loss, val_loss, class_err, roc, auc, conf_mat = func.training(
model, opt.CNN_only, batch_size, epochs, device, optim,
trainset_val, valset)
result_temp = [train_loss, val_loss, class_err, auc]
print("class_err", round(class_err, 3), ", AUC:", round(auc, 3))
results.append(result_temp)
header_results = ['train_loss', 'val_loss', 'class_err', 'auc']
results = pd.DataFrame(results, columns=header_results)
grid_results = pd.concat([grid, results], axis=1)
grid_results = grid_results.sort_values("class_err", ascending=True)
name = "val_results_batch_" + str(opt.grid_nr) + ".csv"
grid_results.to_csv(name, encoding='utf-8', index=False)
print("runtime complete:", datetime.now() - startTime)