Config

class hokohoko.entities.Config(predictor_class: str = 'hokohoko.standard.DoNothing', assessors: list = ['hokohoko.standard.Logger'], data_class: str = 'hokohoko.standard.Npz', data_parameters: str = 'data.npz', data_subset: Optional[str] = None, period_count: int = 256, process_count: int = 8, past_minutes: int = 1440, hold_minutes: int = 1440, load_limit: Optional[int] = None, training_minutes: int = 786240, test_minutes: int = 262080, profiling: bool = False, verbosity: int = 0)

Bases: tuple

Controls the behaviour of Hokohoko. Most of these values should not be changed.

Create new instance of Config(predictor_class, assessors, data_class, data_parameters, data_subset, period_count, process_count, past_minutes, hold_minutes, load_limit, training_minutes, test_minutes, profiling, verbosity)

property predictor_class

The Predictor to use.

property assessors

The Assessor to use.

property data_class

The type of data source to load.

property data_parameters

The data parameters.

property data_subset

Comma-separated list of currency-pair names, or None for all.

property period_count

How many Periods to test.

property process_count

The size of the process Pool (number of cores allocated).

property past_minutes

How many minutes a bar in on_bar represents.

property hold_minutes

How many minutes to ‘Hold’ an Order/Position till auto-close.

property load_limit

How many minutes to make available for testing from. Should be greater than sum(training_minutes, test_minutes, hold_minutes), or None for unlimited.

property training_minutes

How many minutes to train the Predictor on before assessing/accepting trades. Not necessarily contiguous.

property test_minutes

How many minutes to test for. Sequential, but not necessarily contiguous.

property profiling

Enable profiling output in the current directory.

property verbosity

The desired level of debug output.