MetaSweeper Class

Runs all enabled features sweepers.

Load configuration and create sweeper configurations.

Inheritance

builtins.object

MetaSweeper

Constructor

MetaSweeper(task: str, timeout_sec: int = 3600, featurization_config: FeaturizationConfig | None = None, is_cross_validation: bool = False, enable_dnn: bool = False, force_text_dnn: bool = False, feature_sweeping_config: Dict[str, Any] = {})

Parameters

Name	Description
task Required	Task type- Classification, Regression or Forecasting.
timeout_sec	Timeout in seconds for feature sweeping. default value: 3600
is_cross_validation	Whether to do the cross validation. default value: False
feature_sweeping_config	Feature sweeping config. default value: {}
enable_dnn	Flag to enable sweeping over text DNNs such as BERT, BiLSTM. default value: False
force_text_dnn	Flag to force add text DNNs such as BERT, BiLSTM after sweeping. default value: False
featurization_config	default value: None

Methods

sweep

Feature sweeping and / or class balancing sweeping

sweep

Feature sweeping and / or class balancing sweeping

sweep(working_dir: str, X: ndarray | DataFrame | spmatrix | EnginelessDataflow, y: ndarray | Series | Categorical | EnginelessDataflow, stats_and_column_purposes: List[Tuple[RawFeatureStats, str, str]] | None = None, sweeping_mode: str = 'feature') -> List[Any]

Parameters

Name	Description
working_dir Required
X Required
y Required
stats_and_column_purposes	default value: None
sweeping_mode	default value: feature

Attributes

DEFAULT_SWEEPER_TIMEOUT_SEC

DEFAULT_SWEEPER_TIMEOUT_SEC = 3600