ASX Gym Observations
Observations If we ever want to do better than take random actions at each step, it’d probably be good to actually know what our actions are doing to the environment.
The environment’s step function returns exactly what we need. In fact, step returns four values. These are:
- observation (object): an environment-specific object representing your observation of the environment. For example, pixel data from a camera, joint angles and joint velocities of a robot, or the board state in a board game.
- reward (float): the amount of reward achieved by the previous action. The scale varies between environments, but the goal is always to increase your total reward.
- done (boolean): whether it’s time to reset the environment again. Most (but not all) tasks are divided up into well-defined episodes, and done being True indicates the episode has terminated. (For example, perhaps the pole tipped too far, or you lost your last life.)
- info (dict): diagnostic information useful for debugging. It can sometimes be useful for learning (for example, it might contain the raw probabilities behind the environment’s last state change). However, official evaluations of your agent are not allowed to use this for learning.
following is the code snippet for Observations in ASX Gym
self.observation_space = spaces.Dict(
{
"indexes": spaces.Dict({
'open': spaces.Box(low=np.float32(0),
high=np.float32(self.number_infinite),
dtype=np.float32),
'close': spaces.Box(low=np.float32(0),
high=np.float32(self.number_infinite),
dtype=np.float32),
'high': spaces.Box(low=np.float32(0),
high=np.float32(self.number_infinite),
dtype=np.float32),
'low': spaces.Box(low=np.float32(0),
high=np.float32(self.number_infinite),
dtype=np.float32),
}
),
"day": spaces.Discrete(self.number_infinite),
"seconds": spaces.Discrete(24 * 3600),
"company_count": spaces.Discrete(self.max_company_number),
"prices:": spaces.Dict({
"company_id": spaces.MultiDiscrete([self.max_company_number]
* self.max_company_number),
"ask_price": spaces.Box(low=np.float32(0),
high=np.float32(self.max_stock_price),
shape=(self.max_company_number,),
dtype=np.float32),
"bid_price": spaces.Box(low=np.float32(0),
high=np.float32(self.max_stock_price),
shape=(self.max_company_number,),
dtype=np.float32),
"price": spaces.Box(low=np.float32(0),
high=np.float32(self.max_stock_price),
shape=(self.max_company_number,),
dtype=np.float32)}),
"portfolio_company_count": spaces.Discrete(self.max_company_number),
"portfolios": spaces.Dict(
{
"company_id": spaces.MultiDiscrete([self.max_company_number]
* self.max_company_number),
"volume": spaces.Box(np.float32(0),
high=np.float32(self.number_infinite),
shape=(self.max_company_number,),
dtype=np.float32),
"buy_price": spaces.Box(low=np.float32(0),
high=np.float32(self.max_stock_price),
shape=(self.max_company_number,),
dtype=np.float32),
"sell_price": spaces.Box(low=np.float32(0),
high=np.float32(self.max_stock_price),
shape=(self.max_company_number,),
dtype=np.float32),
"price": spaces.Box(low=np.float32(0),
high=np.float32(self.max_stock_price),
shape=(self.max_company_number,),
dtype=np.float32),
}),
"bank_balance:": spaces.Box(low=np.float32(0),
high=np.float32(self.number_infinite),
dtype=np.float32),
"total_value:": spaces.Box(low=np.float32(0),
high=np.float32(self.number_infinite),
dtype=np.float32),
"available_fund:": spaces.Box(low=np.float32(0),
high=np.float32(self.number_infinite),
dtype=np.float32)
}
)what you can see is the daily Stock index, your bank account details, your current state of portfolios.
like action, ASX Gym also provides some helper models to help you process observations
company_id is the id of company ,you can refer to the Company List
class StockIndex:
def __init__(self, index_date, open_index, close_index, high_index, low_index):
self.index_date = index_date
self.open_index = open_index
self.close_index = close_index
self.high_index = high_index
self.low_index = low_index
def to_json_obj(self):
json_obj = {
'index_date': self.index_date,
'open_index': round(self.open_index, 2),
'close_index': round(self.close_index, 2),
'high_index': round(self.high_index, 2),
'low_index': round(self.low_index, 2)
}
return json_obj
class StockPrice:
def __init__(self, price_date, company_id, open_price,
close_price, high_price, low_price):
self.price_date = price_date
self.company_id = company_id
self.open_price = open_price
self.close_price = close_price
self.high_price = high_price
self.low_price = low_price
def to_json_obj(self):
json_obj = {
'price_date': self.price_date,
'company_id': int(self.company_id),
'open_price': round(self.open_price, 2),
'close_price': round(self.close_price, 2),
'high_price': round(self.high_price, 2),
'low_price': round(self.low_price, 2)
}
return json_obj
class StockRecord:
def __init__(self, company_id, volume, buy_price, sell_price, price):
self.company_id = company_id
self.volume = volume
self.buy_price = buy_price
self.sell_price = sell_price
self.price = price
class AsxObservation:
def __init__(self, observation):
self.day = observation['day']
self.seconds = observation['second']
self.total_value = float(observation['total_value'].item())
self.available_fund = float(observation['available_fund'].item())
self.bank_balance = float(observation['bank_balance'].item())
open_index = float(observation['indexes']['open'].item())
close_index = float(observation['indexes']['close'].item())
high_index = float(observation['indexes']['high'].item())
low_index = float(observation['indexes']['low'].item())
self.stock_index = StockIndex('', open_index,
close_index, high_index, low_index)
self.portfolios = []
self.prices = {}
company_count = observation['company_count']
for c in range(company_count):
company_id = observation['prices']['company_id'][c].item()
ask_price = observation['prices']['ask_price'][c].item()
bid_price = observation['prices']['bid_price'][c].item()
price = observation['prices']['price'][c].item()
self.prices[company_id] = StockSimulationPrice(ask_price, bid_price, price)
portfolio_company_count = observation['portfolio_company_count']
for c in range(portfolio_company_count):
company_id = observation['portfolios']['company_id'][c].item()
volume = observation['portfolios']['volume'][c].item()
buy_price = observation['portfolios']['buy_price'][c].item()
sell_price = observation['portfolios']['sell_price'][c].item()
price = observation['portfolios']['price'][c].item()
stock_record = StockRecord(company_id, volume, buy_price, sell_price, price)
self.portfolios.append(stock_record)
def to_json_obj(self):
json_obj = {"day": int(self.day),
"seconds": int(self.seconds),
"total_value": round(self.total_value, 2),
"available_fund": round(self.available_fund, 2),
"bank_balance": round(self.bank_balance, 2),
"index": {
"open": round(self.stock_index.open_index, 2),
"close": round(self.stock_index.close_index, 2),
"high": round(self.stock_index.high_index, 2),
"low": round(self.stock_index.low_index, 2)
},
"prices": {},
"portfolios": {}}
for company_id, prices in self.prices.items():
json_obj["prices"][company_id] = {
"ask_price": round(prices.ask_price, 2),
"bid_price": round(prices.bid_price, 2),
"price": round(prices.price, 2)
}
for stock_record in self.portfolios:
json_obj["portfolios"][stock_record.company_id] = {
"volume": round(stock_record.volume, 2),
"buy_price": round(stock_record.buy_price, 2),
"sell_price": round(stock_record.sell_price, 2),
"price": round(stock_record.price, 2)
}
return json_objsample observation
{
"day":1,
"seconds":36000,
"total_value":100000.0,
"available_fund":96690.03,
"bank_balance":0.0,
"index":{
"open":6746.2,
"close":6735.8,
"high":6759.6,
"low":6735.8
},
"prices":{
2:{
"ask_price":80.91,
"bid_price":80.91,
"price":80.91
},
3:{
"ask_price":40.75,
"bid_price":40.75,
"price":40.75
},
4:{
"ask_price":27.82,
"bid_price":27.82,
"price":27.82
},
5:{
"ask_price":27.0,
"bid_price":27.0,
"price":27.0
},
6:{
"ask_price":27.17,
"bid_price":27.17,
"price":27.17
},
44:{
"ask_price":3.11,
"bid_price":3.11,
"price":3.11
},
67:{
"ask_price":21.84,
"bid_price":21.84,
"price":21.84
},
100:{
"ask_price":61.74,
"bid_price":61.74,
"price":61.74
},
200:{
"ask_price":5.63,
"bid_price":5.63,
"price":5.63
},
300:{
"ask_price":1.75,
"bid_price":1.75,
"price":1.75
}
},
"portfolios":{
200:{
"volume":54.0,
"buy_price":5.63,
"sell_price":0.0,
"price":5.63
},
300:{
"volume":82.0,
"buy_price":1.75,
"sell_price":0.0,
"price":1.75
},
2:{
"volume":23.0,
"buy_price":80.91,
"sell_price":0.0,
"price":80.91
},
4:{
"volume":36.0,
"buy_price":27.82,
"sell_price":0.0,
"price":27.82
}
}
}