skfolio.base.BaseAssetPanelTransformer#

class skfolio.base.BaseAssetPanelTransformer[source]#

Base class for estimators that transform asset panel data.

Descriptors and factor exposure estimators take an AssetPanel and return transformed values indexed by observation and asset. Most transformers return an array with shape (n_observations, n_assets). Transformers that produce multiple values per asset, such as OneHotCategoricalFactors, return an array with shape (n_observations, n_assets, n_categories).

In scikit-learn, fit and partial_fit update fitted state, stored in trailing underscore attributes, while transform returns transformed input data using that state. This separation is not suitable for every AssetPanel transformer. For some estimators, the transformed value is produced by the same state transition that updates the estimator. A separate transform method would either need to mutate state or depend on a preceding partial_fit call, so the API exposes the combined operation directly. For example, the exponentially weighted momentum descriptor EWMomentum needs to update its internal EWMA state to compute the transformed value on each observation.

Other transformers are independent across observations. For example, the DividendToPrice descriptor depends only on the current dividends_ttm and market_cap values and can therefore be declared stateless.

Accordingly, fit_transform is used for full-batch computation and partial_fit_transform for online computation. Subclasses must implement fit_transform. Downstream meta-estimators use the presence of partial_fit_transform to determine whether a transformer supports online transformation.

Supported implementation patterns are:

Batch-only transformers implement only fit_transform.
Stateless transformers declare stateless=True and implement only fit_transform. The base class adds partial_fit_transform as a direct delegation to fit_transform.
Online transformers implement both fit_transform and partial_fit_transform.

Attributes:

n_assets_int: Number of assets seen during fitting.
asset_names_ndarray of shape (n_assets,): Asset names seen during fitting.

Methods

`fit_transform`(X[, y])	Fit the transformer if needed and return transformed values.
`get_metadata_routing`()	Get metadata routing of this object.
`get_params`([deep])	Get parameters for this estimator.
`set_params`(**params)	Set the parameters of this estimator.