Installation

$ pip3 install tsfaker

Simple usage

Generate 3 rows of fake data from a single table schema file.

$ tsfaker https://gitlab.com/healthdatahub/tsfaker/raw/master/tests/schemas/implemented_types.json  --nrows 3 --pretty
  boolean         string            number      integer        date              datetime  year yearmonth
0       1  haHoKysholbSI    9780230269.512  -7061309068  1914-10-03  1902-04-11T11:21:11Z  1939    196405
1       0      rLugGhNek    990894536.8945   2529879443  2026-09-08  2015-11-27T16:21:54Z  1932    192909
2       1         ipqVXm  -4371053960.8987   -529880373  1994-09-27  1937-01-12T18:40:15Z  2021    193303

Advanced usage

Show help message.

$ tsfaker --help
Usage: tsfaker [OPTIONS] [SCHEMA_DESCRIPTORS]...
...

Download examples schemas from project schema-snds.

$ git clone  https://gitlab.com/healthdatahub/schema-snds && cd schema-snds

Generate fake data for all schemas in a schemas folder using csv files in nomenclatures folder, and write them to fake_data folder.

$ mkdir fake_data
$ tsfaker schemas -o fake_data -r nomenclatures
2019-01-01 00:00:00 :: INFO :: Data generated from descriptor 'schemas/PMSI/PMSI MCO/T_MCOaa_nnE.json' will be written on 'fake_data/PMSI/PMSI MCO/T_MCOaa_nnE.csv'
2019-01-01 00:00:00 :: INFO :: Data generated from descriptor 'schemas/PMSI/PMSI MCO/T_MCOaa_nnFASTC.json' will be written on 'fake_data/PMSI/PMSI MCO/T_MCOaa_nnFASTC.csv'
2019-01-01 00:00:00 :: INFO :: Data generated from descriptor 'schemas/PMSI/PMSI SSR/T_SSRaa_nnE.json' will be written on 'fake_data/PMSI/PMSI SSR/T_SSRaa_nnE.csv'
...

Simple data Generators

• numpy comes with many functions to generate random data.

• rstr and exrex generate random string following regular expressions.

• Faker and Mimesis allow to generate fake data. They both focus on generating high level data, such as names, email or addresses, which does not seem necessary for us.

• DataScienceFaker generate synthetic data conforming to statistical distributions. It is based on numpy and rstr.

Table generator

• pydbgen is a shallow wrapper around Faker to generate tables as pandas dataframe, sqlite table or Excel files.

• pySyntheticDatasetGenerator is a wrapper around dsfaker, that generate tables with their relations as described in yaml configuration files.

• datafiller generate random data from database schema. API could be interesting.

• plaitpy is a fake table generator from a yaml configuration file.

Realistic data

Generating realistic data - ie data carrying statistical information - could mean different things in different contexts :

• realistic statistical distribution on single columns,

• realistic temporal dynamics,

• realistic correlations between pairs of columns,

• realistic correlations between pairs of columns from different (joinable) tables,

• etc.

Hence there is no universal way to generate realistic data. Most approaches follow two steps :

1. learn a statistical model from the real data,

2. generate data using this model.

The statistical model depends of the context, and is usually not expressed in the form of a generic schema, such as table-schema. However, a schema of your data will be often be necessary to configure this kind of libraries.

This topic is an active research area, with many articles but few production implementations :

• DataSynthesizer (article) learn a diferentially private Bayesian network capturing the correlation structure between attributes

• dpgan (article) Differentially Private Releasing via Deep Generative Model.

• SDV (article) Generative modeling for relational databases.

• medGAN (article) Generative adversarial network for generating electronic health records.

• CTGAN (article) Modeling Tabular data using Conditional GAN

The statistical model may convey sensitive information and personnal data. It is important fact to bear in mind, as protecting sensitive information is a common reason to generate fake data in the first place.

Some tools offer ways to mitigate the risk from personal data leakage, with no formal guarantees. Other tools offer formal privacy guarantees through differential privacy.

An active line of work is to use Generative Adversial Network to generate realistic data, for example dpgan (see above) or Privacy-Preserving Generative Deep Neural Networks Support Clinical Data Sharing.

When using Neural Network, one can use TensorFlow’s specific library. PySyft project aims to provide a generic implementation for PyTorch.