Skip to main content

Safety net for machine learning pipelines.

Project description

redflag

Tests Docs PyPI version Conda Version PyPI versions PyPI license

🚩 redflag aims to be an automatic safety net for machine learning datasets. The vision is to accept input of a Pandas DataFrame or NumPy ndarray representing the input X and target y in a machine learning task. redflag will provide an analysis of each feature, and of the target, including aspects such as class imbalance, leakage, outliers, anomalous data patterns, threats to the IID assumption, and so on. The goal is to complement other projects like pandas-profiling and greatexpectations.

Installation

You can install this package with pip:

python -m pip install redflag

Alternatively, you can use the conda package manager, pointed at the conda-forge channel:

conda install -c conda-forge redflag

For developers, there is a pip option for installing dev dependencies. Use pip install "redflag[dev]" to install all testing and documentation packages.

Example with sklearn

The most useful components of redflag are probably the scikit-learn "detectors". These sit in your pipeline, look at your training and validation data, and emit warnings if something looks like it might cause a problem. For example, we can get alerted to an imbalanced target vector y like so:

import redflag as rf
from sklearn.datasets import make_classification

X, y = make_classification(weights=[0.1])

_ = rf.ImbalanceDetector().fit(X, y)

This raises a warning:

🚩 The labels are imbalanced by more than the threshold (0.780 > 0.400). See self.minority_classes_ for the minority classes.

For maximum effect, put this and other detectors in your pipeline, or use the pre-build rf.pipeline which contains several useful alerts.

See the documentation, and specifically the notebook Using redflag with sklearn.ipynb for other examples.

Example of function call

redflag is also a collection of functions. Most of the useful ones take one or more columns of data (usually a 1D or 2D NumPy array) and run a single test. For example, we can do some outlier detection. The get_outliers() function returns the indices of data points that are considered outliers:

>>> import redflag as rf
>>> data = 3 * [-3, -2, -2, -1, 0, 0, 0, 1, 2, 2, 3]
>>> rf.get_outliers(data)
array([], dtype=int64)

That is, there are no outliers. But let's add a clear outlier: a new data record with a value of 100. The function returns the index position(s) of the outlier point(s):

>>> rf.get_outliers(data + [100])
array([33])

See the documentation, and specifically the notebook Basic_usage.ipynb for several other basic examples.

Documentation

The documentation is online.

Contributing

Please see CONTRIBUTING.md. There is also a section in the documentation about Development.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

redflag-0.5.0.tar.gz (1.2 MB view hashes)

Uploaded Source

Built Distribution

redflag-0.5.0-py3-none-any.whl (53.6 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page