A library to build and test machine learning features
This library provides a set of tools that can be useful in many machine learning applications (classification, clustering, regression, etc.), and particularly helpful if you use scikit-learn (although this can work if you have a different algorithm).
Most machine learning problems involve an step of feature definition and preprocessing. Feature Forge helps you with:
- Defining and documenting features
- Testing your features against specified cases and against randomly generated cases (stress-testing). This helps you making your application more robust against invalid/misformatted input data. This also helps you checking that low-relevance results when doing feature analysis is actually because the feature is bad, and not because there’s a slight bug in your feature code.
- Evaluating your features on a data set, producing a feature evaluation matrix. The evaluator has a robust mode that allows you some tolerance both for invalid data and buggy features.
- Experimentation: running, registering, classifying and reproducing experiments for determining best settings for your problems.
pip install featureforge.
Documentation is available at http://feature-forge.readthedocs.org/en/latest/
Feature Forge is © 2014 Machinalis (http://www.machinalis.com/). Its primary authors are:
Any contributions or suggestions are welcome, the official channel for this is submitting github pull requests or issues.
- Bug fixes related to sparse matrices.
- Small documentation improvements.
- Reduced default logging verbosity.
- Using sparse numpy matrices by default.
- Discarded the need of using forked version of Schema library.
- Added support for running and generating stats for experiments
- Fixing installer dependencies
- Added support for python 3
- Added support for bag-of-words features
- Initial release