Skip to main content

Dataset synthesis for Python

Project description

PySynth: Dataset Synthesis for Python

PySynth is a package to create synthetic datasets - that is, datasets that look just like the original in terms of statistical properties, variable values, distributions and correlations, but do not have exactly the same contents so are safe against data disclosure. An alternative to R's Synthpop with a more permissive license.

Installation

You can get PySynth from PyPI by using the obvious

pip install pysynth

Usage

You can perform the synthesis with basic settings directly on a CSV file:

python -m pysynth source.csv synthesized.csv

This produces a synthesized.csv file that will look a lot like the original (variable names values, distributions, correlations) but will (most likely) not be the same.

For better control, it is best to use the synthesizer objects. They follow the scikit-learn interface for Pandas dataframes so you fit() them on the original and then sample(n) to get a synthetic dataframe of n rows.

So far, only a synthesizer based on iterative proportional fitting (pysynth.ipf.IPFSynthesizer) is available. This synthesis bins continuous variables to categories and reconstructs them using fitted univariate distributions. Missing values (NaN) are preserved.

Synthesis quality measurement modules to be added.

Contributors

Feedback, additions, suggestions, issues and pull requests are welcome and much appreciated on GitHub.

How to add features:

  1. Fork it (https://github.com/simberaj/pysynth/fork)
  2. Create your feature branch (git checkout -b feature/feature-name)
  3. Commit your changes (git commit -am "feature-name added")
  4. Push to the branch (git push origin feature/feature-name)
  5. Create a new pull request

Development requires pytest for testing and sphinx to generate documentation. Tests can be run using simple

pytest tests

Intended development directions

  • Synthesis quality measurement in terms of anonymization/similarity
  • Model-based synthesis along the lines of R's Synthpop

License and author info

PySynth is developed by Jan Šimbera simbera.jan@gmail.com.

PySynth is available under the MIT license. See LICENSE.txt for more details.

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

pysynth-0.0.4.tar.gz (13.7 kB view details)

Uploaded Source

Built Distribution

pysynth-0.0.4-py3-none-any.whl (14.7 kB view details)

Uploaded Python 3

File details

Details for the file pysynth-0.0.4.tar.gz.

File metadata

  • Download URL: pysynth-0.0.4.tar.gz
  • Upload date:
  • Size: 13.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.6.0 requests-toolbelt/0.9.1 tqdm/4.40.2 CPython/3.7.5

File hashes

Hashes for pysynth-0.0.4.tar.gz
Algorithm Hash digest
SHA256 996703b8a3585dd766e79a6969d25d4cc76f921461f903cf15165d59ed23f6d5
MD5 254c50cdf541018608b009e2e4582e88
BLAKE2b-256 96a7852abd24b6e894ebf9fbc44bbbc14a145bcfef07d14bc8fb7f8bc8e2086d

See more details on using hashes here.

File details

Details for the file pysynth-0.0.4-py3-none-any.whl.

File metadata

  • Download URL: pysynth-0.0.4-py3-none-any.whl
  • Upload date:
  • Size: 14.7 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.1.1 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.6.0 requests-toolbelt/0.9.1 tqdm/4.40.2 CPython/3.7.5

File hashes

Hashes for pysynth-0.0.4-py3-none-any.whl
Algorithm Hash digest
SHA256 a00853471996ddc3513a621155007d998a046c202691b57ab9b76eed2a0db2b1
MD5 61ce1272d3cf13834dbef9c785e12a64
BLAKE2b-256 6d1faa1587287d24aaefb88a098fb2aa027a7ea019bd1576bb16565fac71fdd5

See more details on using hashes here.

Supported by

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