def prepare_data(self):
df = pd.read_csv(self.input_data_file)
df = df.drop(['Symbol'], axis=1)
df = df.sort_values(['Date'])
df = add_indicators(df.reset_index())
train_len = int(len(df) * 0.8)
df = df[:train_len]
validation_len = int(train_len * 0.8)
self.train_df = df[:validation_len]
self.test_df = df[validation_len:]
def __init__(self, df, initial_balance=10000, commission=0.0003, reward_func='sortino', **kwargs):
super(BitcoinTradingEnv, self).__init__()
self.initial_balance = initial_balance
self.commission = commission
self.reward_func = reward_func
self.df = df.fillna(method='bfill')
self.df = add_indicators(self.df.reset_index())
self.stationary_df = log_and_difference(
self.df, ['Open', 'High', 'Low', 'Close', 'Volume BTC', 'Volume USD'])
self.n_forecasts = kwargs.get('n_forecasts', 10)
self.confidence_interval = kwargs.get('confidence_interval', 0.95)
self.obs_shape = (1, 5 + len(self.df.columns) -
2 + (self.n_forecasts * 3))
# Actions of the format Buy 1/4, Sell 3/4, Hold (amount ignored), etc.
self.action_space = spaces.Discrete(12)
# Observes the price action, indicators, account action, price forecasts
self.observation_space = spaces.Box(
low=0, high=1, shape=self.obs_shape, dtype=np.float16)
def getDatasets(inputfile,
testPercentage=20,
percentageToUse=100,
withBug=True):
df = pd.read_csv(inputfile)
df = df.drop(['Symbol'], axis=1)
if not withBug:
df['Date'] = pd.to_datetime(
df['Date'], format='%Y-%m-%d %I-%p'
) # parse date from string to the correct format
df = df.sort_values(['Date'])
df = add_indicators(df.reset_index())
# Just to speed things up
print('> Using {}% of the dataset'.format(percentageToUse))
df = df[:int(len(df) * percentageToUse / 100)]
test_len = int(len(df) * testPercentage / 100)
train_len = int(len(df)) - test_len
train_df = df[:train_len]
test_df = df[train_len:]
return train_df, test_df
curr_idx = -1
reward_strategy = 'sortino'
input_data_file = 'data/coinbase_hourly.csv'
params_db_file = 'sqlite:///params.db'
study_name = 'ppo2_' + reward_strategy
study = optuna.load_study(study_name=study_name, storage=params_db_file)
params = study.best_trial.params
print("Training PPO2 agent with params:", params)
print("Best trial reward:", -1 * study.best_trial.value)
df = pd.read_csv(input_data_file)
df = df.drop(['Symbol'], axis=1)
df = df.sort_values(['Date'])
df = add_indicators(df.reset_index())
test_len = int(len(df) * 0.2)
train_len = int(len(df)) - test_len
train_df = df[:train_len]
test_df = df[train_len:]
train_env = DummyVecEnv([lambda: BitcoinTradingEnv(
train_df, reward_func=reward_strategy, forecast_len=int(params['forecast_len']), confidence_interval=params['confidence_interval'])])
test_env = DummyVecEnv([lambda: BitcoinTradingEnv(
test_df, reward_func=reward_strategy, forecast_len=int(params['forecast_len']), confidence_interval=params['confidence_interval'])])
model_params = {
'n_steps': int(params['n_steps']),
curr_idx = 3
reward_strategy = 'sortino'
input_data_file = 'binance.csv'
params_db_file = 'sqlite:///params.db'
study_name = 'ppo2_' + reward_strategy
study = optuna.load_study(study_name=study_name, storage=params_db_file)
params = study.best_trial.params
print("Testing PPO2 agent with params:", params)
print("Best trial:", -1 * study.best_trial.value)
df_init = pd.read_csv('binance.csv')
# df = df.drop(['Symbol'], axis=1)
df_init = df_init.sort_values(['Date'])
df_init = add_indicators(df_init.reset_index())
test_len = int(len(df_init) * 0.021)
train_len = int(len(df_init)) - test_len
# test_df = df_init[train_len:]
# test_df = df_init[16400:]
# print('before load test_env')
# test_env = DummyVecEnv([lambda: BitcoinTradingEnv(
# test_df, reward_func=reward_strategy, forecast_len=int(params['forecast_len']), confidence_interval=params['confidence_interval'])])
# print('after load test_env')
# test_env = DummyVecEnv([lambda: BitcoinTradingEnv(
# test_df, reward_func=reward_strategy, forecast_len=4, confidence_interval=0.81)])
# model_params = {
# 'n_steps': int(params['n_steps']),
