base_environment

Classes

BaseTradingEnv

Base trading environment for reinforcement learning agents.

Module Contents

class base_environment.BaseTradingEnv(data, cfg, features, time_step=(TimeFrameUnit.Day, 1))

Bases: gymnasium.Env, abc.ABC

Base trading environment for reinforcement learning agents. Contains common functionality shared between training and testing environments.

Parameters:

  • data (pandas.DataFrame)

  • cfg (trading.cli.alg.config.StockEnv)

  • features (list[trading.src.features.generic_features.Feature] | list[str])

  • time_step (tuple[alpaca.data.timeframe.TimeFrameUnit, int])

symbols
stock_dimension
features
feature_cols
cfg: trading.cli.alg.config.StockEnv
time_step
action_space
observation_space
observation_index
terminal = False
observation_timestamp = None
init_data(data)

Set the data for the environment. :param data: DataFrame containing the trading data.

Parameters:

data (pandas.DataFrame)

classmethod observation(df, portfolio_state, feature_cols, prices)

Build observation from dataframe, portfolio state, and prices.

Parameters:

  • df (pandas.DataFrame)

  • portfolio_state (numpy.ndarray)

  • feature_cols (list)

  • prices (numpy.ndarray)

Return type:

numpy.ndarray

_get_observation_df(i=-1)

Get the dataframe slice for observation at index i.

Parameters:

i (int)

Return type:

pandas.DataFrame

_get_prices(i=-1)

Get prices at index i.

Parameters:

i (int)

Return type:

numpy.ndarray

abstractmethod _get_observation(i=-1)

Get the current observation from the environment.

Parameters:

i (int)

Return type:

numpy.ndarray

reset(*, seed=None, options=None)

Reset the environment to its initial state. Can be overridden by subclasses.

Parameters:

  • seed (Optional[int])

  • options (Optional[dict])

abstractmethod step(action)

Execute one time step within the environment.

render()

Render the environment state.

_reset_internal_states(timestamp=None)

Reset internal state counters.

Parameters:

timestamp (pandas.Timestamp | None)