About the project

Skoot is designed to provide as much flexibility as possible while offering implementations to common challenges, such as categorical & model-based imputation transformers, transformers to rectify skewness (i.e., box-cox & Yeo-Johnson transformations), as well as many wrappers to scikit-learn transformers that enable applications to selected columns only. Every transformer in skoot is designed for maximum flexibility and to minimize impact on existing pipelines. Each transformer has been tested to function in the scope of scikit-learn pipelines and grid searches, and offers the same persistence model.

In addition, skoot provides transformer profiling utilities to help you identify bottlenecks, determine whether a model will perform sufficiently in production, and help you make any other performance-based considerations.

from skoot.preprocessing import BoxCoxTransformer
from skoot.preprocessing import SelectiveStandardScaler
from skoot.utils.profiling import profile_estimator
from sklearn.datasets import load_iris
from sklearn.pipeline import Pipeline
from sklearn.linear_model import LogisticRegression

# load data
X, y = load_iris(return_X_y=True)

# fit a pipeline
pipe = Pipeline([
    ('transform', BoxCoxTransformer(n_jobs=2)),
    ('scale', SelectiveStandardScaler(cols=[0, 1])),
    ('clf', LogisticRegression())
]).fit(X, y)

# profile
print(profile_estimator(pipe))

Which yields:

[('transform', (('fit_time_', 0.1410069465637207),)), ('scale', (('fit_time_', 0.0011677742004394531),))]