SDV (Synthetic Data Vault)
Open-source ecosystem for generating synthetic tabular, multi-table, and time-series data.
About
The Synthetic Data Vault (SDV) is an open-source ecosystem for synthetic data generation and evaluation. It was created at MIT's Data to AI Lab in 2016 and published as an IEEE DSAA 2016 paper by Neha Patki, Roy Wedge, and Kalyan Veeramachaneni; DataCebo was founded in 2020 to grow the project and is its developer.
SDV bundles libraries for data discovery and reversible transforms, multiple synthesizer models ranging from statistical copulas to deep learning (such as GaussianCopulaSynthesizer), and SDMetrics for measuring the quality and privacy of synthetic datasets. It supports single-table, multi-table relational, and time-series synthesis through a metadata-driven fit-and-sample API, installs with pip install sdv, and DataCebo offers an enterprise edition with air-gapped deployment.
What it can do
- Single-table, multi-table, and time-series synthesis
- Synthesizer models from copulas to deep learning
- Metadata-driven fit and sample API
- SDMetrics for quality and privacy evaluation
- Enterprise edition with air-gapped deployment
Setup
pip install sdv (or conda install -c pytorch -c conda-forge sdv); load a demo dataset and fit a synthesizer in a few lines of Python.
Prerequisites
Python 3.9+; a real dataset whose statistical patterns the synthesizer should learn.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 94; Trending score 80/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.