An open source Python library for automated feature engineering based on Genetic Programming.
- Free software: BSD license
- Documentation: https://evolutionary-forest.readthedocs.io
Feature engineering is a long-standing issue that has plagued machine learning practitioners for many years. Deep learning techniques have significantly reduced the need for manual feature engineering in recent years. However, a critical issue is that the features discovered by deep learning methods are difficult to interpret.
In the domain of interpretable machine learning, genetic programming has demonstrated to be a promising method for automated feature construction, as it can improve the performance of traditional machine learning systems while maintaining similar interpretability. Nonetheless, such a potent method is rarely mentioned by practitioners. We believe that the main reason for this phenomenon is that there is still a lack of a mature package that can automatically build features based on the genetic programming algorithm. As a result, we propose this package with the goal of providing a powerful feature construction tool for enhancing existing state-of-the-art machine learning algorithms, particularly decision-tree-based algorithms.
- A powerful feature construction tool for generating interpretable machine learning features.
- A reliable machine learning model with powerful performance on small datasets.
pip install -U evolutionary_forestAn example of usage:
X, y = load_diabetes(return_X_y=True)
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)
r = EvolutionaryForestRegressor(max_height=3, normalize=True, select='AutomaticLexicase',
gene_num=10, boost_size=100, n_gen=20, n_pop=200, cross_pb=1,
base_learner='Random-DT', verbose=True)
r.fit(x_train, y_train)
print(r2_score(y_test, r.predict(x_test)))An example of improvements brought about by constructed features:
Here are some notebook examples of using Evolutionary Forest:
Tutorial: English Version | 中文版本
This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.