sktools provides tools to extend sklearn, like several feature engineering based transformers.

To install sktools, run this command in your terminal:

$ pip install sktools

Can be found in https://sktools.readthedocs.io

from sktools import IsEmptyExtractor

from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import Pipeline

...

mod = Pipeline([
    ("impute-features", IsEmptyExtractor()),
    ("model", LogisticRegression())
])

...

Here's a list of features that sktools currently offers:

• sktools.encoders.NestedTargetEncoder performs target encoding suited for variables with nesting.
• sktools.encoders.QuantileEncoder performs target aggregation using a quantile instead of the mean.
• sktools.preprocessing.CyclicFeaturizer converts numeric to cyclical features via sine and cosine transformations.
• sktools.impute.IsEmptyExtractor creates binary variables indicating if there are missing values.
• sktools.matrix_denser.MatrixDenser transformer that converts sparse matrices to dense.
• sktools.quantilegroups.GroupedQuantileTransformer creates quantiles of a feature by group.
• sktools.quantilegroups.PercentileGroupFeaturizer creates features regarding how an instance compares with a quantile of its group.
• sktools.quantilegroups.MeanGroupFeaturizer creates features regarding how an instance compares with the mean of its group.
• sktools.selectors.TypeSelector gets variables matching a type.
• sktools.selectors.ItemsSelector allows to manually choose some variables.
• sktools.ensemble.MedianForestRegressor applies the median instead of the mean when aggregating trees predictions.
• sktools.linear_model.QuantileRegression sklearn style wrapper for quantile regression.
• sktools.model_selection.BootstrapFold bootstrap cross-validator.
• sktools.GradientBoostingFeatureGenerator Automated feature generation through gradient boosting.

Contributing --------

Fork/clone, in a fresh environment, run:

$ pip install -e ".[dev]"

To check if the unit tests are ok, run

$ make test

MIT license

This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.