scikit-lego 0.3.0

a collection of lego bricks for scikit-learn pipelines

scikit-lego

We love scikit learn but very often we find ourselves writing custom transformers, metrics and models. The goal of this project is to attempt to consolidate these into a package that offers code quality/testing. This project is a collaboration between multiple companies in the Netherlands.

Note that we're not formally affiliated with the scikit-learn project at all. Same holds with lego.

Installation

Install scikit-lego via pip with

pip install scikit-lego

Alternatively, to edit and contribute you can fork/clone and run:

pip install -e ".[dev]"
python setup.py develop

Documentation

The documentation can be found here.

Usage

# the scikit learn stuff we love
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline

# from scikit lego stuff we add
from sklego.preprocessing import RandomAdder
from sklego.mixture import GMMClassifier

...

mod = Pipeline([
    ("scale", StandardScaler()),
    ("random_noise", RandomAdder()),
    ("model", GMMClassifier())
])

...

Features

Here's a list of features that this library currently offers:

• sklego.preprocessing.PatsyTransformer applies a patsy formula
• sklego.preprocessing.RandomAdder adds randomness in training
• sklego.preprocessing.PandasTypeSelector selects columns based on pandas type
• sklego.preprocessing.ColumnSelector selects columns based on column name
• sklego.preprocessing.ColumnCapper limits extreme values of the model features
• sklego.preprocessing.OrthogonalTransformer makes all features linearly independant
• sklego.dummy.RandomRegressor benchmark that predicts random values
• sklego.naive_bayes.GaussianMixtureNB classifies by training a 1D GMM per column per class
• sklego.mixture.GMMClassifier classifies by training a GMM per class
• sklego.mixture.GMMOutlierDetector detects outliers based on a trained GMM
• sklego.pipeline.DebugPipeline adds debug information to make debugging easier
• sklego.meta.DecayEstimator adds decay to the sample_weight that the model accepts
• sklego.meta.GroupedEstimator can split the data into runs and run a model on each
• sklego.meta.EstimatorTransformer adds a model output as a feature
• sklego.metrics.correlation_score calculates correlation between model output and feature
• sklego.metrics.p_percent_score proxy for model fairness with regards to sensitive attribute
• sklego.datasets.load_chicken loads in the joyful chickweight dataset
• sklego.datasets.make_simpleseries make a simulated timeseries
• sklego.pandas_utils.log_step a simple logger-decorator for pandas pipeline steps
• sklego.pandas_utils.add_lags adds lag values of certain columns in pandas

New Features

We want to be rather open here in what we accept but we do demand three things before they become added to the project:

1. any new feature contributes towards a demonstratable real-world usecase
2. any new feature passes standard unit tests (we have a few for transformers and predictors)
3. the feature has been discussed in the issue list beforehand