def test_catches_config_error(self, config_filename, excpecetd_exception):
"""Checks if data collection catches configuration errors.
Args:
config_filename: the filename for the config file.
excpecetd_exception: the expected exception.
"""
with self.assertRaises(excpecetd_exception):
create_data_collection(read_config_file(config_filename))
def test_checks_date_range_set(self):
"""Checks if data collection catches when date range is not set.
"""
config = read_config_file("test/data_collection.yaml")
data_collection = create_data_collection(config)
with self.assertRaises(ValueError):
data_collection.prepare_data()
def test_catches_circular_dependency(self):
"""Checks if data collection catches circular dependency.
"""
config = read_config_file(
"test/circular_dependency_data_collection.yaml")
data_collection = create_data_collection(config)
with self.assertRaises(ValueError):
data_collection.reset()
def test_use_relative(self):
"""Checks if data collection uses relative stock data properly.
"""
config = read_config_file("test/data_collection.yaml")
config["use_relative_stock_data"] = True
data_collection = create_data_collection(config)
self.assertEqual(4, len(data_collection.data_objects))
self.assertNotEqual(data_collection.stock_data_id,
data_collection.absolute_stock_data_id)
def test_ignores_duplicates(self):
"""Checks if data collection ignores duplicates.
"""
config = read_config_file("test/data_collection.yaml")
data_collection = create_data_collection(config)
stock_data = RealStockData()
returned_stock_data = data_collection.append(stock_data)
self.assertIsInstance(returned_stock_data, RealStockData)
self.assertNotEqual(stock_data, returned_stock_data)
def test_get_availbale_dates(self):
"""Checks if get available dates works properly.
"""
from_date = datetime(2016, 1, 1)
to_date = datetime(2016, 2, 1)
config = read_config_file("test/data_collection.yaml")
data_collection = create_data_collection(config)
data_collection.set_date_range(from_date, to_date)
data_collection.prepare_data()
available_dates = data_collection.get_available_dates()
self.assertEqual(10, len(available_dates))
def test_adds_randomization(self):
"""Checks if data collection adds randomization properly.
"""
config = read_config_file("test/data_collection.yaml")
config["stock_data_randomization"] = True
data_collection = create_data_collection(config)
self.assertEqual(3, len(data_collection.data_objects))
config = read_config_file("test/simulation.yaml")
config["stock_data_randomization"] = False
data_collection = create_data_collection(config)
self.assertEqual(1, len(data_collection.data_objects))
config = read_config_file("test/simulation.yaml")
config["stock_data_randomization"] = True
data_collection = create_data_collection(config)
self.assertEqual(2, len(data_collection.data_objects))
randomization_layer = data_collection.data_objects[0]
stock_data = data_collection.data_objects[1]
self.assertEqual(randomization_layer.dependencies[0],
stock_data.id_str)
def _add_id_with_hash_values(self):
"""Adds hash values to the id_str
"""
data_collection = create_data_collection(self.data_collection_config)
reward = None
model = None
if not self.reward_config is None:
reward = create_reward(self.reward_config, None, None)
if not self.model_config is None:
model = create_model(self.model_config)
self.id_str_with_hash = "{}_{}_{}_{}".format(self.id_str,
hash(data_collection),
hash(reward), hash(model))
def test_prepares_data(self):
"""Checks if data collection prepares the data properly.
"""
from_date = datetime(2016, 1, 1)
to_date = datetime(2016, 2, 1)
config = read_config_file("test/data_collection.yaml")
data_collection = create_data_collection(config)
self.assertEqual(data_collection.stock_data_id,
data_collection.absolute_stock_data_id)
data_collection.set_date_range(from_date, to_date)
data_collection.prepare_data()
for data in data_collection.data_objects:
self.assertTrue(data.ready)
def test_getitem(self):
"""Checks if __getitem__ works as expected.
"""
expected_index = [
"GOOG", "AMZN", "ra_10_stock_data_GOOG", "ra_10_stock_data_AMZN"
]
from_date = datetime(2016, 1, 1)
to_date = datetime(2016, 2, 1)
config = read_config_file("test/data_collection.yaml")
data_collection = create_data_collection(config)
data_collection.set_date_range(from_date, to_date)
data_collection.prepare_data()
available_dates = data_collection.get_available_dates()
self.assertTrue((expected_index == data_collection[
available_dates[0]].index.tolist()))
def test_resets_data(self):
"""Checks if data collection prepares the data properly.
"""
from_date = datetime(2016, 1, 1)
to_date = datetime(2016, 2, 1)
config = read_config_file("test/data_collection.yaml")
data_collection = create_data_collection(config)
data_collection.set_date_range(from_date, to_date)
data_collection.prepare_data()
for data in data_collection.data_objects:
self.assertTrue(data.ready)
self.assertEqual(3, data_collection.recursive_counter)
data_collection.reset()
for data in data_collection.data_objects:
if data.name != "real_stock_data":
self.assertFalse(data.ready)
else:
self.assertTrue(data.ready)
data_collection.prepare_data()
self.assertEqual(2, data_collection.recursive_counter)
def test_creates_data_collection(self):
"""Checks if creates data collection.
"""
data_collection = create_data_collection(
read_config_file("test/data_collection.yaml"))
self.assertIsInstance(data_collection, DataCollection)
def __init__(self, data_collection_config=None, from_date=None, to_date=None, min_duration=0,
max_duration=0, min_start_balance=1000, max_start_balance=1000, commission=0,
max_stock_owned=1, stock_data_randomization=False, reward_config=None):
"""Initializer for the simulation class.
Args:
data_collection_config: configuration of the data configuration.
from_date: datetime date for the start of the range
to_date: datetime date for the end of the range
min_duration: minimum length of the episode.
max_duration: maximum length of the episode (if 0 will run for all available dates).
min_start_balance: minimum starting balance.
max_start_balance: maximum starting balance. Balance selected unifromly.
commission: relative commission for each transcation.
max_stock_owned: a maximum number of different stocks that can be owned.
stock_data_randomization: whether to add stock data randomization.
reward_config: the configuration for the reward.
"""
if data_collection_config is None:
data_collection_config = read_config_file(DEFAULT_DATA_COLLECTION_CONFIG_FILE)
data_collection_config["stock_data_randomization"] = stock_data_randomization
self.data_collection_config = data_collection_config
if from_date is None and to_date is None:
from_date = datetime(2014, 1, 1)
to_date = datetime(2016, 1, 1)
elif from_date is None or to_date is None:
raise ValueError("Either both from and to dates are None or none of them.")
self.data_collection = create_data_collection(data_collection_config)
# Setup of simulation data.
stock_data_id = self.data_collection.absolute_stock_data_id
self.stock_data = self.data_collection.id_to_data[stock_data_id]
self.stock_names = self.data_collection.stock_names
self.balance = 0
self.net_worth = 0
self.owned_stocks = None
# Adding buffer days.
buffer = self.data_collection.get_buffer()
from_date -= timedelta(days=buffer)
# Setting date range for data collection.
self.data_collection.set_date_range(from_date, to_date)
self.data_collection.prepare_data()
self.available_dates = self.data_collection.get_available_dates()
# Setting duration range.
max_duration = min(max_duration, len(self.available_dates))
self.max_duration = max_duration if max_duration > 0 else len(self.available_dates)
self.min_duration = min_duration if min_duration > 0 else self.max_duration
self.min_duration = min(self.max_duration, self.min_duration)
# Setting starting balance range.
self.min_start_balance = min_start_balance
self.max_start_balance = max_start_balance
# Setting date tracking.
self.curr_date_index = -1
self.from_date_index = -1
self.to_date_index = -1
# Setting commission and max stock owned.
self.commission = commission
self.max_stock_owned = max_stock_owned
# Setting up action space.
self.action_space = (
gym.spaces.MultiDiscrete([3] * len(self.data_collection.stock_names))
)
# Setting up reward function.
if reward_config is None:
reward_config = {"name": "net_worth_ratio_reward"}
self.reward_function = create_reward(reward_config, from_date, to_date)
# Setting up observation cache.
self.saved_date_index = -1
self.saved_observation = None
def test_hash(self):
"""Checks if __hash__ works.
"""
config = read_config_file("test/data_collection.yaml")
data_collection = create_data_collection(config)
self.assertIsNotNone(hash(data_collection))
def test_get_buffer(self):
"""Checks if buffer days is calculated correctly.
"""
config = read_config_file("test/data_collection.yaml")
data_collection = create_data_collection(config)
self.assertEqual(10, data_collection.get_buffer())