# h5 savefile
if not os.path.exists(args.out_dir):
os.makedirs(args.out_dir)
train_store = utils.create_hdf5(args.out_dir+'/train_'+args.start_day+'_'+args.end_day)
validate_store = utils.create_hdf5(args.out_dir+'/validate_'+args.start_day+'_'+args.end_day)
test_store = utils.create_hdf5(args.out_dir+'/test_'+args.start_day+'_'+args.end_day)
# save dates to file
train_f = open(args.out_dir+'/datelist_train_'+args.start_day+'_'+args.end_day+'.txt', 'w')
validate_f = open(args.out_dir+'/datelist_validate_'+args.start_day+'_'+args.end_day+'.txt', 'w')
test_f = open(args.out_dir+'/datelist_test_'+args.start_day+'_'+args.end_day+'.txt', 'w')
utils.save_dates_to_file(train_f, train_set_str)
utils.save_dates_to_file(validate_f, validate_set_str)
utils.save_dates_to_file(test_f, test_set_str)
train_set_str = utils.gen_lkbk_days(day_list=train_set)
validate_set_str = utils.gen_lkbk_days(day_list=validate_set)
test_set_str = utils.gen_lkbk_days(day_list=test_set)
else:
# Pick a date at random
# Generate a list of 60 business days starting from the random date chosen
start_day = sample(trading_days, 1)[0]
if args.chosen_day is not None:
start_day = datetime.strptime(args.chosen_day, '%Y%m%d')
training_set = date_range(start_day, periods=args.ndays, freq='B')
train_set_str = [date.date().strftime('%Y%m%d') for date in training_set]
# h5 savefile
if not os.path.exists(args.out_dir):
os.makedirs(args.out_dir)
train_store = utils.create_hdf5(args.out_dir+'/'+start_day.strftime('%Y%m%d'))
training_set_str = [line[:-1] for line in f]
training_set = [
datetime.strptime(x, '%Y%m%d').replace(hour=9, minute=30)
for x in training_set_str
]
# file to backup console prints
log_file = open(args.out_dir + '/log_' + args.dataset[:-3] + '.txt', 'w')
today_data_all = {}
lkbk_days_data_all = {}
multiplier = {}
for i in range(len(training_set)):
today = training_set[i]
lkbk_days = utils.gen_lkbk_days(today=today)
lkbk_days = [datetime.strptime(x, '%Y%m%d') for x in lkbk_days]
utils._print(log_file, 'processing %s' % today)
try:
today_time_range = utils.day_time_range(today)
vols = input_comp.ix[:, today, 'Volume']
closes = input_comp.ix[:, today, 'Close']
liqs = vols * closes
liqs = liqs / liqs.sum()
liqs = liqs.fillna(0)
# get top input components by liquidity
liqs = liqs[np.argsort(liqs)[::-1]]
topn_input = liqs.index[:]
# fetch/generate dates
f = open(args.in_dir+'/datelist_'+args.dataset[:-3]+'.txt')
training_set_str = [line[:-1] for line in f]
training_set = [datetime.strptime(x, '%Y%m%d').replace(hour=9, minute=30) for x in training_set_str]
# file to backup console prints
log_file = open(args.out_dir+'/log_'+args.dataset[:-3]+'.txt', 'w')
today_data_all = {}
lkbk_days_data_all = {}
multiplier = {}
for i in range(len(training_set)):
today = training_set[i]
lkbk_days = utils.gen_lkbk_days(today=today)
lkbk_days = [datetime.strptime(x,'%Y%m%d') for x in lkbk_days]
utils._print(log_file, 'processing %s' % today)
try:
today_time_range = utils.day_time_range(today)
vols = input_comp.ix[:, today, 'Volume']
closes = input_comp.ix[:, today, 'Close']
liqs = vols * closes
liqs = liqs / liqs.sum()
liqs = liqs.fillna(0)
# get top input components by liquidity
liqs = liqs[np.argsort(liqs)[::-1]]
topn_input = liqs.index[:]
# save dates to file
train_f = open(
args.out_dir + '/datelist_train_' + args.start_day + '_' +
args.end_day + '.txt', 'w')
validate_f = open(
args.out_dir + '/datelist_validate_' + args.start_day + '_' +
args.end_day + '.txt', 'w')
test_f = open(
args.out_dir + '/datelist_test_' + args.start_day + '_' +
args.end_day + '.txt', 'w')
utils.save_dates_to_file(train_f, train_set_str)
utils.save_dates_to_file(validate_f, validate_set_str)
utils.save_dates_to_file(test_f, test_set_str)
train_set_str = utils.gen_lkbk_days(day_list=train_set)
validate_set_str = utils.gen_lkbk_days(day_list=validate_set)
test_set_str = utils.gen_lkbk_days(day_list=test_set)
else:
# Pick a date at random
# Generate a list of 60 business days starting from the random date chosen
start_day = sample(trading_days, 1)[0]
if args.chosen_day is not None:
start_day = datetime.strptime(args.chosen_day, '%Y%m%d')
training_set = date_range(start_day, periods=args.ndays, freq='B')
train_set_str = [date.date().strftime('%Y%m%d') for date in training_set]
# h5 savefile
if not os.path.exists(args.out_dir):
os.makedirs(args.out_dir)
train_store = utils.create_hdf5(args.out_dir + '/' +