Release Notes#

sklearndf 2.3#

2.3.0#

sklearndf 2.3 adds support for scikit-learn 1.3 and drops support for scikit-learn 0.24.

API: add DF wrapper classe HDBSCANDF for native estimator HDBSCAN
API: add DF wrapper class TargetEncoderDF for native estimator TargetEncoder

sklearndf 2.2#

sklearndf 2.2 adds support for scikit-learn 1.2, and enhances the EstimatorDF API.

2.2.1#

VIZ: use scikit-learn’s native HTML representation of estimators, if available

2.2.0#

sklearndf 2.2 adds support for scikit-learn 1.2. It drops support for scikit-learn 0.23 and earlier due to incomplete support of sparse output (see below).

API: DF estimators now support native estimators using sparse matrices as input or output, and automatically convert them to or from sparse DataFrame objects
API: new property EstimatorDF.output_names_ to get the names of the output columns the estimator was fitted with
API: new method LearnerPipelineDF.preprocess to apply the preprocessing step to a data frame
API: remove properties feature_names_out_ and feature_names_original_ from class LearnerPipelineDF
API: Index instances obtained from EstimatorDF.feature_names_in_ and TransformerDF.feature_names_out_ are now named "feature" instead of "feature_in" and "feature_out", respectively, and Series instances obtained from TransformerDF.feature_names_original_ are now named "feature_original" instead of "feature_in", and their indices are now named "feature" instead of "feature_out"; this is to separate the semantics of the originating property from the column index, which may be used in other contexts

sklearndf 2.1#

sklearndf 2.1 adds support for scikit-learn 1.1.

2.1.1#

This is a maintenance release to catch up with sklearndf 2.0.2.

2.1.0#

API: new clusterer BisectingKMeansDF
API: new transformer MiniBatchNMFDF
API: new transformer RandomTreesEmbeddingDF; note that class RandomTreesEmbedding existed previously in scikit-learn, but is based on TransformerMixin only as of scikit-learn 1.1
API: support parameters max_categories and min_frequency of OneHotEncoderDF, introduced in scikit-learn 1.1
API: support array-like values for parameter drop of OneHotEncoderDF
API: support "passthrough" as a transformer in FeatureUnionDF
API: remove GeneralizedLinearRegressorDF since the underlying native estimator is a base class and not intended to be used as a regressor of its own

sklearndf 2.0#

sklearndf 2.0 adds support for scikit-learn 1.0, adds data frame support for clusterers along with additional API enhancements and improvements, and is now subject to static type checking with mypy.

2.0.2#

BUILD: add support for pandas 2.0 and above
FIX: property PCADF.n_components_ now returns the value of n_components_, not n_components
FIX: detect missing and extra columns when validating data frames resulting from transforms, even when the total column count is correct

2.0.1#

API: upon declaration of new wrapper classes, automatically validate that their associated native estimators are compatible with the wrapper class
API: new public constants DROP and PASSTHROUGH in ColumnTransformerDF
FIX: base LGBMClassifierDF and XGBClassifierDF on the the correct wrapper class ClassifierWrapperDF
FIX: support array-like values for parameter drop of OneHotEncoderDF
FIX: various minor tweaks and stability improvements

2.0.0#

API: ClassifierDF and RegressorDF get a new base class SupervisedLearnerDF, which in turn is based on LearnerDF; SupervisedLearnerDF implements method score(), which is no longer implemented by LearnerDF
API: new class ClusterDF, based on LearnerDF
API: class EstimatorDF now implements the HasExpressionRepr mix-in, rendering estimator representations as Expression objects to enable better formatting
API: added data frame support for method partial_fit()
API: removed OutlierRemoverDF
API: removed dependency on package lightgbm: LGBMClassifierDF and LGBMRegressorDF are still available if lightgbm is installed
API: added support for xgboost: XGBClassifierDF and XGBClassifierDF are available if xgboost is installed
API: DF wrapper classes are now created using proper class declarations to better conform with Python type conventions checked by mypy; see sklearndf.wrapper for details
API: remove functions make_df_estimator, make_df_classifier, make_df_regressor, and make_df_transformer which are now obsolete
API: move some classes in sklearndf.wrapper to sub-packages sklearndf.wrapper.stacking and sklearndf.wrapper.numpy to improve package navigability and to achieve better de-coupling of the underlying code; this change also moves StackingClassifierWrapperDF and StackingRegressorWrapperDF to package sklearndf.wrapper.stacking

sklearndf 1.2#

This release adds support for scikit-learn 0.24.

1.2.3#

This is a maintenance release to catch up with sklearndf 1.1.3.

1.2.2#

This release makes small API tweaks, and catches up with sklearndf 1.1.2.

API: make type hints more specific in signatures for make_df_transformer(), make_df_classifier(), and make_df_regressor()

1.2.1#

This is a maintenance release to catch up with sklearndf 1.1.1.

1.2.0#

API: add DF adaptations for classes introduced by scikit-learn 0.24: PolynomialCountSketchDF and SequentialFeatureSelectorDF

sklearndf 1.1#

1.1.3#

This release relaxes package dependencies to support any numpy version 1.x from 1.16.

1.1.2#

This release improves compatibility with scikit-learn and fixes bugs.

API: add full support for the _estimator_type attribute
FIX: do not reset transformers when calling TransformerDF.inverse_transform()
FIX: accept “passthrough” as value for arg remainder of ColumnTransformerDF

1.1.1#

This release addresses compatibility issues with meta-estimators.

FIX: support complex DF estimators inside StackingEstimatorDF
FIX: raise an exception if a base estimator is not supported by one of sklearndf’s implementations for DF meta-estimators

1.1.0#

This release exposes the wrapper API used to generate augmented DF estimators from native scikit-learn estimators.

API: expose the EstimatorWrapperDF class hierarchy through the new sklearndf.wrapper package
API: create new scikit-learn wrapper classes with the new functions make_df_estimator(), make_df_classifier(), make_df_regressor(), and make_df_transformer()

sklearndf 1.0#

1.0.2#

This is a maintenance release focusing on enhancements to the CI/CD pipeline and bug fixes.

FIX: correctly mirror __init__ signatures of native estimators to their corresponding DF estimators
FIX: do not mirror native estimator class attributes and protected members to DF estimators
FIX: support "passthrough" transformer in ColumnTransformerDF
FIX: support drop parameter in OneHotEncoderDF
BUILD: add support for numpy 1.20
BUILD: updates and changes to the CI/CD pipeline

1.0.1#

Initial release.

FAQ